1 00:00:00,060 --> 00:00:02,440 Michael: Hello and welcome to Postgres.FM, a weekly show about 2 00:00:02,440 --> 00:00:03,300 all things PostgreSQL. 3 00:00:03,560 --> 00:00:05,240 I am Michael, founder of pgMustard. 4 00:00:05,280 --> 00:00:06,800 This is Nikolay, founder of Postgres.AI. 5 00:00:07,200 --> 00:00:09,440 Hey Nikolay, what are we talking about this week? 6 00:00:09,960 --> 00:00:14,940 Nikolay: Hi Michael, I thought sometimes it's a good idea to 7 00:00:15,060 --> 00:00:17,280 reconsider decisions, you know. 8 00:00:18,620 --> 00:00:23,860 And if you chose some tool or some project to work with, sometimes 9 00:00:23,860 --> 00:00:30,920 it's good to look back at recent experience and further back 10 00:00:31,340 --> 00:00:35,820 in the past, why you made this decision and think is it still 11 00:00:35,860 --> 00:00:36,720 the right decision? 12 00:00:37,640 --> 00:00:43,220 And should you keep using this product or system or project or 13 00:00:43,220 --> 00:00:47,060 anything or tool or it's time to change it? 14 00:00:47,980 --> 00:00:50,140 Michael: Yeah, I hope you're not having second thoughts. 15 00:00:51,660 --> 00:00:52,940 Nikolay: You can doubt anything. 16 00:00:53,940 --> 00:00:54,660 Why not? 17 00:00:55,960 --> 00:00:59,220 Michael: Just I had on my list of people I expected to doubt 18 00:00:59,220 --> 00:01:04,760 things Nikolay Samokhvalov doubting Postgres was very, very low 19 00:01:04,760 --> 00:01:05,700 down my list. 20 00:01:06,820 --> 00:01:10,120 Nikolay: Okay, so, by the way, good pronunciation of my last 21 00:01:10,120 --> 00:01:10,620 name. 22 00:01:11,760 --> 00:01:12,260 Yes. 23 00:01:12,340 --> 00:01:13,580 Yeah, thank you. 24 00:01:14,540 --> 00:01:19,940 And yeah, so, honestly, it's not a joke. 25 00:01:20,800 --> 00:01:23,800 I'm thinking, should I still use Postgres? 26 00:01:25,640 --> 00:01:29,120 And for me, of course, it's a much more fundamental question, 27 00:01:30,780 --> 00:01:32,280 rather than for like... 28 00:01:33,080 --> 00:01:37,480 Compared to people even if you even if you CTO and all your data 29 00:01:37,480 --> 00:01:40,180 in Postgres it's a big question for you to... 30 00:01:41,240 --> 00:01:44,480 But for me it's even a bigger question because a company called 31 00:01:44,480 --> 00:01:52,700 Postgres.AI and last 19 years I put all data to Postgres, last... 32 00:01:53,300 --> 00:01:55,300 Okay, how many years? 33 00:01:56,000 --> 00:01:59,400 Postgres consultants, like, a source of income depends on it. 34 00:01:59,800 --> 00:02:03,040 Not only data in Postgres, but also all my activities are in 35 00:02:03,040 --> 00:02:03,480 Postgres. 36 00:02:03,480 --> 00:02:10,640 And since 2007, I'm also a community activities, having some 37 00:02:10,640 --> 00:02:16,940 community activities, all my thoughts related to professional 38 00:02:17,860 --> 00:02:24,240 stuff very close to Postgres, around Postgres, but I must admit 39 00:02:24,240 --> 00:02:25,080 I have doubts. 40 00:02:26,400 --> 00:02:31,100 So what I would like to for us today discussing our topic which 41 00:02:31,100 --> 00:02:32,100 is why Postgres. 42 00:02:34,140 --> 00:02:39,480 What I would like to have like doing this discussion is to I 43 00:02:39,480 --> 00:02:43,220 would like to look inside myself, raise all the doubts I have, 44 00:02:43,740 --> 00:02:49,860 and honestly, If the doubts grow and we don't resolve them, I'm 45 00:02:49,860 --> 00:02:56,320 ready to switch to a different database system and a different 46 00:02:56,320 --> 00:02:57,040 type of business. 47 00:02:57,040 --> 00:02:57,940 I'm not joking. 48 00:02:59,760 --> 00:03:01,200 Michael: I'm looking forward to this. 49 00:03:01,800 --> 00:03:04,100 Nikolay: I see huge potential recently. 50 00:03:04,160 --> 00:03:07,800 We started growing in all aspects and it's super interesting. 51 00:03:07,800 --> 00:03:10,780 and so many things not yet accomplished. 52 00:03:12,180 --> 00:03:17,360 I have big ideas, I think I have big ideas too, which I must 53 00:03:17,360 --> 00:03:18,740 implement and so on. 54 00:03:18,740 --> 00:03:20,180 But I also have doubts. 55 00:03:20,240 --> 00:03:29,880 Let's talk about them and if they start to overweigh these ideas 56 00:03:30,220 --> 00:03:36,640 and desire to build and so on, then I'm honestly open to switch. 57 00:03:37,660 --> 00:03:38,160 Really. 58 00:03:38,240 --> 00:03:42,520 It's also probably I won't be present on this podcast anymore. 59 00:03:42,520 --> 00:03:43,420 I'm not joking. 60 00:03:44,840 --> 00:03:45,340 Really. 61 00:03:45,420 --> 00:03:47,780 So, why Postgres? 62 00:03:47,780 --> 00:03:48,480 Michael: Let's go back. 63 00:03:48,480 --> 00:03:49,320 Let's go back. 64 00:03:49,340 --> 00:03:50,700 Should we start at the beginning? 65 00:03:52,080 --> 00:03:56,380 Why did we each choose Postgres initially, both as users, but 66 00:03:56,380 --> 00:04:01,400 also to build tools around, build companies around, build careers 67 00:04:01,420 --> 00:04:01,920 around? 68 00:04:02,140 --> 00:04:04,540 It'd be interesting to hear why you chose it, even if it's... 69 00:04:04,540 --> 00:04:07,120 I know we've covered it briefly before, but... 70 00:04:07,120 --> 00:04:09,100 Nikolay: You want to start from me or from yourself? 71 00:04:10,080 --> 00:04:11,740 Michael: Why not start from you, I think? 72 00:04:12,340 --> 00:04:13,160 Nikolay: Sounds good. 73 00:04:13,440 --> 00:04:15,900 I went to a great university in 1998. 74 00:04:17,180 --> 00:04:19,500 I think elite, like very good. 75 00:04:20,340 --> 00:04:22,460 I'm an MIT, it's in Russia. 76 00:04:23,040 --> 00:04:27,240 It's like hard to get into 1 of the best technical universities. 77 00:04:29,440 --> 00:04:34,640 And applied maths and computer science was focus, but smaller 78 00:04:34,640 --> 00:04:37,700 focus starting third grade you need to choose specialty. 79 00:04:38,680 --> 00:04:44,240 I wanted to do hardcore stuff so I went to kernel and operational 80 00:04:44,260 --> 00:04:48,520 systems and so on, but I remember it was a couple of Armenian guys 81 00:04:48,520 --> 00:04:54,440 who were professors that were responsible for this direction. 82 00:04:54,860 --> 00:04:59,080 I went to them, they said, write us a letter, we will respond. 83 00:04:59,080 --> 00:05:00,560 And it's great. 84 00:05:00,860 --> 00:05:04,000 So I wrote a letter, no response, wrote a letter once again, 85 00:05:04,000 --> 00:05:04,600 no response. 86 00:05:04,600 --> 00:05:09,940 So in my head I was thinking, damn, I tried to be hardcore. 87 00:05:10,080 --> 00:05:16,660 Now I'm like, I approached guys who are older than me, have experience 88 00:05:16,880 --> 00:05:20,040 in the same institute, in the same subdivision. 89 00:05:21,760 --> 00:05:26,260 Many directions, by the way, AI was already one of the directions, 90 00:05:26,760 --> 00:05:30,040 very different than today, of course, but still, like neural 91 00:05:30,040 --> 00:05:31,580 networks and so on. 92 00:05:31,800 --> 00:05:34,020 Somehow it wasn't interested in that direction. 93 00:05:34,060 --> 00:05:36,500 I personally was asked like what's the easiest? 94 00:05:37,300 --> 00:05:42,540 And they said, oh, professor for like which is doing databases 95 00:05:44,020 --> 00:05:49,080 is the easiest And he's like so kind, you know, like, and definitely 96 00:05:49,200 --> 00:05:50,140 it's the easiest. 97 00:05:50,460 --> 00:05:51,660 Like, good. 98 00:05:52,060 --> 00:05:52,920 I go there. 99 00:05:52,920 --> 00:05:54,360 This is how I chose databases. 100 00:05:55,580 --> 00:05:56,080 Honestly. 101 00:05:57,180 --> 00:06:00,100 Then I started saying, you know, databases are in the heart of 102 00:06:00,100 --> 00:06:02,420 any information system. 103 00:06:02,440 --> 00:06:03,840 It's where data is stored. 104 00:06:03,840 --> 00:06:05,740 It should be in good shape. 105 00:06:05,740 --> 00:06:13,220 If it's in bad shape, then everything can be bad, slow, and so 106 00:06:13,220 --> 00:06:13,720 on. 107 00:06:14,280 --> 00:06:19,640 Not efficient or not bad quality, like not secure or something. 108 00:06:19,640 --> 00:06:21,780 So many things, it's really in the center, right? 109 00:06:21,780 --> 00:06:22,940 I believe in this. 110 00:06:23,080 --> 00:06:26,300 But at that time I chose just like by accident. 111 00:06:27,440 --> 00:06:33,900 So, and then I worked with Oracle slightly more than 1 year, 112 00:06:33,960 --> 00:06:38,440 with SQL Server maybe up to a couple of years, and then somehow 113 00:06:38,440 --> 00:06:44,120 by accident they got into startups in 2004-5, and they said of 114 00:06:44,120 --> 00:06:48,060 course we cannot buy, we have like budget to build something. 115 00:06:48,900 --> 00:06:54,180 I joined and became CTO, like started growing team and I was 116 00:06:54,180 --> 00:06:58,240 responsible for almost all decisions technically and they said 117 00:06:58,240 --> 00:07:00,720 we cannot afford Oracle license of course. 118 00:07:00,760 --> 00:07:04,900 I was a big fan of SQL Server and I told them, okay, then we 119 00:07:04,900 --> 00:07:08,600 use the most popular open source database system, which is MySQL. 120 00:07:09,520 --> 00:07:13,200 And you know, I probably mentioned, right, 1 week later I was 121 00:07:13,200 --> 00:07:19,880 already almost quitting because it's ridiculous, repair table, 122 00:07:20,360 --> 00:07:22,040 0000, date. 123 00:07:22,200 --> 00:07:23,780 I cannot work with it. 124 00:07:23,780 --> 00:07:29,440 I have good foundation, because professor was kind and easy going, 125 00:07:29,440 --> 00:07:34,600 but it also was great professor, and the best I think in Russia 126 00:07:34,600 --> 00:07:38,240 at that time in terms of academia and very well known books and 127 00:07:38,240 --> 00:07:38,620 so on. 128 00:07:38,620 --> 00:07:43,500 I was very lucky to be his student and then PhD student and his 129 00:07:43,500 --> 00:07:44,880 name is Sergei Kuznetsov. 130 00:07:45,900 --> 00:07:48,640 His name was unfortunately he died recently. 131 00:07:50,360 --> 00:07:52,580 So yeah, he was a very great guy. 132 00:07:53,320 --> 00:07:58,320 And I learned a lot from him and of course I learned, I studied 133 00:07:58,360 --> 00:08:04,960 books like Chris Date's, like kind of simple thing, but also 134 00:08:04,960 --> 00:08:07,440 Garcia-Molina, Ullman, right? 135 00:08:08,300 --> 00:08:13,440 Which is… Later I learned these guys are from Stanford. 136 00:08:13,980 --> 00:08:15,560 It was unexpected. 137 00:08:15,600 --> 00:08:18,760 I thought maybe it should be from Berkeley or so, but then I 138 00:08:18,760 --> 00:08:25,100 also found they are already not active but from Stanford. 139 00:08:25,580 --> 00:08:32,720 So many good stuff like relational theory, calculus, These things. 140 00:08:33,480 --> 00:08:37,200 Sometimes I learned many things in Russian, so sometimes I'm 141 00:08:37,200 --> 00:08:40,500 having difficulties finding proper English names. 142 00:08:41,040 --> 00:08:43,400 But in general, I had a good foundation. 143 00:08:43,840 --> 00:08:47,960 SQL at that time already learned like, it's actually from him 144 00:08:47,960 --> 00:08:50,940 I learned that nulls is a big problem in SQL model. 145 00:08:51,060 --> 00:08:54,400 There is a relational model and there is a SQL model and they 146 00:08:54,400 --> 00:08:56,140 are not completely the same. 147 00:08:56,380 --> 00:08:59,740 At that time I studied also SQL standard. 148 00:09:00,240 --> 00:09:05,360 I remember SQL 92, then 1999, then they started doing this 2.0.0.m 149 00:09:06,180 --> 00:09:12,280 or x, like kind of not whole big release, but releasing parts, 150 00:09:12,280 --> 00:09:12,780 right? 151 00:09:13,060 --> 00:09:16,540 So a lot of stuff learned there. 152 00:09:18,060 --> 00:09:20,660 Standard is hard to deal with for sure. 153 00:09:20,660 --> 00:09:24,440 So I like big respect to guys who deal with it. 154 00:09:25,360 --> 00:09:27,460 Peter Eisentrout and Vic Fearing, right? 155 00:09:28,480 --> 00:09:28,660 CB. 156 00:09:28,660 --> 00:09:29,670 From the Postgres side. 157 00:09:29,670 --> 00:09:29,760 LB. 158 00:09:29,760 --> 00:09:32,420 The closest to standard, these 2 guys, right? 159 00:09:32,780 --> 00:09:33,540 Big respect. 160 00:09:33,940 --> 00:09:35,300 For sure, it's hard. 161 00:09:35,740 --> 00:09:41,760 So I was very young, like 20, 21, 22, I was already reading the 162 00:09:41,760 --> 00:09:42,100 standard. 163 00:09:42,100 --> 00:09:46,460 Then I work with MySQL and what a piece of something it is, actually, 164 00:09:46,460 --> 00:09:46,960 honestly. 165 00:09:47,580 --> 00:09:51,300 We learned some principles, but you touch something and it doesn't 166 00:09:51,300 --> 00:09:55,120 work or work in opposite direction than you compared to what 167 00:09:55,120 --> 00:09:56,240 you learned in Oracle. 168 00:09:57,180 --> 00:10:02,280 In SQL Server, of course, there are some specifics, but in general, 169 00:10:02,720 --> 00:10:07,700 both systems were mature and worked quite well and definitely 170 00:10:07,860 --> 00:10:09,140 like grown-up systems. 171 00:10:09,140 --> 00:10:11,020 MySQL wasn't a grown-up system. 172 00:10:11,160 --> 00:10:13,580 You need to choose at that time, right? 173 00:10:13,580 --> 00:10:17,920 You need to choose if it's MyISAM, repair table, but it lacks 174 00:10:17,920 --> 00:10:23,260 full-text search and all you need to do was the second engine. 175 00:10:23,260 --> 00:10:24,740 So… CB. 176 00:10:24,960 --> 00:10:25,460 Michael: NDB? 177 00:10:25,760 --> 00:10:26,260 LB. 178 00:10:26,280 --> 00:10:28,440 Nikolay: NDB, right, but it's super slow. 179 00:10:29,760 --> 00:10:30,660 Super slow. 180 00:10:31,160 --> 00:10:37,180 So 1 or 2 weeks I was on this, but we just hired a guy from astronomical 181 00:10:37,440 --> 00:10:42,100 university from the same place where Oleg Bartunov was, I think 182 00:10:42,100 --> 00:10:46,400 is still doing some academia activities. 183 00:10:46,880 --> 00:10:48,340 Maybe already know, I don't know. 184 00:10:48,340 --> 00:10:49,220 But at that time... 185 00:10:49,900 --> 00:10:54,260 So that guy said, I know this guy, Oleg Bartunov, and try Postgres. 186 00:10:55,240 --> 00:10:58,880 And everything went in the right place immediately. 187 00:10:59,140 --> 00:11:01,240 I converted super fast everything. 188 00:11:01,620 --> 00:11:02,420 Super fast. 189 00:11:02,920 --> 00:11:07,540 I needed to write an additional layer of caching because I had my 190 00:11:07,540 --> 00:11:09,220 own ORM at that time. 191 00:11:09,220 --> 00:11:13,240 It was before I think current popular ORMs became popular. 192 00:11:13,380 --> 00:11:14,840 It was before Django. 193 00:11:15,180 --> 00:11:19,540 I think Django and Ruby on Rails already existed, but were not 194 00:11:19,540 --> 00:11:22,160 super popular, and we were using PHP. 195 00:11:24,760 --> 00:11:29,920 I wrote my own and it had a lot of work with metadata. 196 00:11:31,120 --> 00:11:34,040 I quickly switched to Postgres and worked with metadata in Postgres, 197 00:11:34,040 --> 00:11:35,560 but it became very slow. 198 00:11:35,580 --> 00:11:38,540 Everything is good, but working with metadata is much slower 199 00:11:38,800 --> 00:11:44,600 than in MySQL, so I needed to add caching for queries to pg_class 200 00:11:44,720 --> 00:11:45,600 and so on. 201 00:11:46,080 --> 00:11:47,800 But this is the only concern. 202 00:11:48,180 --> 00:11:51,580 Everything else went like it was like all the pieces of the puzzle 203 00:11:51,580 --> 00:11:54,020 in the same, like where they should be, right? 204 00:11:55,440 --> 00:12:00,040 Because Postgres probably has issues like replication, like we 205 00:12:00,040 --> 00:12:04,100 learned later and needed to use Slony and Londiste later. 206 00:12:04,540 --> 00:12:10,520 But it works as expected if you studied databases from books, 207 00:12:10,520 --> 00:12:11,020 right? 208 00:12:12,040 --> 00:12:18,240 From theory, Switching from theory to Postgres is natural, which 209 00:12:18,240 --> 00:12:21,800 was not so if you switch from theory to MySQL. 210 00:12:22,120 --> 00:12:28,100 This is how I end up actually loving Postgres because it saved 211 00:12:28,100 --> 00:12:28,600 me. 212 00:12:29,200 --> 00:12:32,980 It's free because we couldn't afford license being a small startup. 213 00:12:33,680 --> 00:12:34,440 It's open. 214 00:12:35,740 --> 00:12:39,660 Open maybe I didn't really care that much at that time about 215 00:12:39,660 --> 00:12:42,140 openness, but I liked it. 216 00:12:43,380 --> 00:12:46,160 Extensibility, openness, I liked it immediately. 217 00:12:46,720 --> 00:12:51,300 And most importantly for me, things work as expected in terms 218 00:12:51,300 --> 00:12:57,540 of relational theory, like there are triggers, there are views, 219 00:12:58,620 --> 00:13:04,120 right, and like null behavior and so on, like these things as 220 00:13:04,140 --> 00:13:05,640 you expect them to be. 221 00:13:06,180 --> 00:13:07,420 Really reliable, right? 222 00:13:07,420 --> 00:13:13,280 The ACID principles, strict system was already very strict to 223 00:13:13,280 --> 00:13:14,360 ACID principles. 224 00:13:14,440 --> 00:13:19,540 It probably was not super performant at that time compared to 225 00:13:19,540 --> 00:13:20,580 MySQL MyISAM. 226 00:13:21,340 --> 00:13:23,600 But you didn't lose data. 227 00:13:25,580 --> 00:13:26,940 Except it's a bug, right? 228 00:13:26,940 --> 00:13:28,240 Bugs happen, of course. 229 00:13:30,640 --> 00:13:31,140 Michael: Yeah. 230 00:13:31,560 --> 00:13:34,300 So that was originally why you chose Postgres. 231 00:13:34,340 --> 00:13:38,080 And I guess we've gone a little bit into why you continue to 232 00:13:38,080 --> 00:13:40,780 choose it in terms of you fell in love with it. 233 00:13:40,920 --> 00:13:45,080 But why, yeah, why do you continue to build products and services 234 00:13:45,100 --> 00:13:45,860 for it? 235 00:13:46,220 --> 00:13:47,720 Nikolay: Well, this is my personal. 236 00:13:48,160 --> 00:13:49,800 I built 3 social networks. 237 00:13:50,460 --> 00:13:52,560 Well, I was in Russia then. 238 00:13:53,480 --> 00:14:00,720 12 years ago, migrated to the US and lost my interest in social 239 00:14:00,720 --> 00:14:01,600 networks already. 240 00:14:02,280 --> 00:14:06,000 But interest to databases was constant all the time. 241 00:14:06,580 --> 00:14:10,580 Yes, and Postgres popularity grew, demand started to grow, I 242 00:14:10,580 --> 00:14:15,060 think, in 2014-15, and this is exactly when I needed additional 243 00:14:15,060 --> 00:14:18,840 source of income, so I started helping other companies which 244 00:14:19,860 --> 00:14:23,580 started having bigger databases in Postgres, like Postgres databases, 245 00:14:23,580 --> 00:14:25,280 and started having troubles. 246 00:14:25,660 --> 00:14:31,320 It was natural for me to use my knowledge and desire to work 247 00:14:31,320 --> 00:14:36,840 with Postgres and that interest into databases to start a consulting 248 00:14:36,980 --> 00:14:42,500 practice and then this podcast was born out of these seeds, meetups 249 00:14:42,500 --> 00:14:48,340 in the past and desire to go online with meetup activity because 250 00:14:48,340 --> 00:14:52,440 I know people are everywhere, like, and meetups are only in 1 251 00:14:52,440 --> 00:14:52,900 place. 252 00:14:52,900 --> 00:14:55,020 It's good to meet in person, but yeah. 253 00:14:55,080 --> 00:14:58,680 So it was quite natural, everything was natural. 254 00:14:59,440 --> 00:15:01,160 So why Postgres? 255 00:15:01,260 --> 00:15:05,080 Because it's just, I have an interest, I have the opportunity, things 256 00:15:05,080 --> 00:15:08,040 like match, and we continue working with it, right? 257 00:15:09,780 --> 00:15:10,720 Michael: Yeah, perfect. 258 00:15:11,100 --> 00:15:12,320 I'm only hearing positives. 259 00:15:16,360 --> 00:15:20,200 Should I go through mine quickly, the reasons for choosing it 260 00:15:20,200 --> 00:15:20,700 initially? 261 00:15:25,680 --> 00:15:28,840 I think some of them are quite different from yours, but it's also 262 00:15:28,840 --> 00:15:29,340 later. 263 00:15:29,640 --> 00:15:30,860 My journey is later. 264 00:15:30,860 --> 00:15:34,960 So I first came across Postgres through work rather than through 265 00:15:34,960 --> 00:15:36,360 academia or anything. 266 00:15:36,660 --> 00:15:37,900 My degree was unrelated. 267 00:15:38,920 --> 00:15:43,740 So I ended up coming across it first while I was working at a 268 00:15:43,740 --> 00:15:44,980 database tools company. 269 00:15:45,120 --> 00:15:50,780 So I first fell in love with software and it was almost entirely 270 00:15:50,860 --> 00:15:54,500 graphical user interfaces to things like databases. 271 00:15:55,520 --> 00:16:00,400 So our biggest divisions at the company were for SQL Server tools. 272 00:16:00,720 --> 00:16:02,360 So that was my first experience. 273 00:16:02,360 --> 00:16:06,360 And I came across Postgres first as something people started 274 00:16:06,360 --> 00:16:06,860 requesting. 275 00:16:07,100 --> 00:16:09,640 Support, yeah, can you do the same tools? 276 00:16:09,720 --> 00:16:12,600 We'd love to buy the same tools from you for Postgres. 277 00:16:12,640 --> 00:16:17,300 You know, companies that used and loved our tools for SQL Server 278 00:16:17,440 --> 00:16:22,080 were starting to do some projects on Postgres, you know, moving 279 00:16:22,260 --> 00:16:26,040 not that many migrations at the time, but like greenfield projects, 280 00:16:26,040 --> 00:16:27,380 they would often choose Postgres. 281 00:16:27,380 --> 00:16:31,220 So this was about that same time that that 2013, 2014, 282 00:16:31,400 --> 00:16:31,900 2015. 283 00:16:32,380 --> 00:16:36,140 So it's those years that I started to hear it more and more. 284 00:16:36,280 --> 00:16:42,240 And then in late 2014, I actually moved to a startup in London, 285 00:16:43,220 --> 00:16:46,660 who were a payments company, our payments company GoCardless, 286 00:16:46,960 --> 00:16:51,760 and their team, it was exceptional engineering team, still is. 287 00:16:53,940 --> 00:16:55,160 And they love Postgres. 288 00:16:55,520 --> 00:16:59,180 And that really, like, hearing that bigger companies were starting 289 00:16:59,180 --> 00:17:04,600 to use it, and then also knowing that the cool YC startup scene 290 00:17:04,600 --> 00:17:08,360 were also using Postgres, that really triggered something in 291 00:17:08,360 --> 00:17:08,480 me. 292 00:17:08,480 --> 00:17:10,600 It's like, ah, this isn't just something. 293 00:17:10,600 --> 00:17:13,220 Like, I already knew that it was up and coming. 294 00:17:13,260 --> 00:17:16,080 I know it's really old at the same time, but I already knew it 295 00:17:16,080 --> 00:17:17,300 was becoming more popular. 296 00:17:17,720 --> 00:17:22,580 And that these engineers that were much younger and VC-backed 297 00:17:22,800 --> 00:17:27,520 startup in the UK chose it and loved it really inspired me to 298 00:17:27,520 --> 00:17:32,000 like, think this could be a platform I'd want to build for this. 299 00:17:32,000 --> 00:17:33,320 This could be going places. 300 00:17:33,320 --> 00:17:34,060 And also. 301 00:17:34,940 --> 00:17:39,360 It's clearly a technology that is already like sound and fundamentally 302 00:17:39,560 --> 00:17:42,660 sound if the, all of these companies are choosing it. 303 00:17:42,780 --> 00:17:47,080 And I'd also really kind of reacted negatively to some of the 304 00:17:47,080 --> 00:17:49,280 other things that had been hyped up at a similar time. 305 00:17:49,280 --> 00:17:54,800 Like I really couldn't get behind the whole MongoDB, like the 306 00:17:54,800 --> 00:17:56,760 NoSQL movement at the time. 307 00:17:56,760 --> 00:17:58,440 And also, do you remember big data? 308 00:17:58,440 --> 00:18:00,300 Do you remember that as like a phase? 309 00:18:01,320 --> 00:18:06,140 Just really reacted strongly, like strongly and negatively towards 310 00:18:06,140 --> 00:18:08,320 a lot of the people that were hyping those up because it just 311 00:18:08,320 --> 00:18:12,420 didn't seem based on solid foundation, like on good principles. 312 00:18:12,920 --> 00:18:15,800 And it didn't make sense to me that these things were needed 313 00:18:15,800 --> 00:18:17,980 or like fundamentally better. 314 00:18:19,540 --> 00:18:23,480 And so when I saw Postgres was coming, was up and coming, and 315 00:18:23,480 --> 00:18:26,640 it was based on good fundamentals, and they prioritised, like 316 00:18:26,640 --> 00:18:29,080 I saw a lot of value alignment, like you, 317 00:18:29,240 --> 00:18:29,440 Nikolay: I 318 00:18:29,440 --> 00:18:33,800 Michael: would, And I think like most people, correctness and 319 00:18:33,800 --> 00:18:36,500 reliability are more important than performance. 320 00:18:36,840 --> 00:18:40,120 Of course, performance matters after that, but there's no point 321 00:18:40,120 --> 00:18:41,760 in it being fast and wrong. 322 00:18:42,340 --> 00:18:46,560 It's just that priority order seems so much better in the Postgres 323 00:18:46,560 --> 00:18:47,060 ecosystem. 324 00:18:47,240 --> 00:18:49,900 But then beyond that, there were loads of other benefits. 325 00:18:49,900 --> 00:18:54,940 Like as you say, free as in price, that's obviously extremely 326 00:18:54,960 --> 00:18:55,460 attractive. 327 00:18:55,520 --> 00:19:00,040 And I think probably still underappreciated, like actually free, 328 00:19:00,040 --> 00:19:01,420 like totally free. 329 00:19:02,260 --> 00:19:05,220 And then also the fact that it's open from a source code perspective. 330 00:19:06,220 --> 00:19:10,060 Having built tools for SQL Server and Oracle, having been involved 331 00:19:10,080 --> 00:19:14,060 in teams, having to research new versions or look back to why 332 00:19:14,060 --> 00:19:16,400 things have broken, all sorts of things. 333 00:19:16,400 --> 00:19:20,280 If you don't have access to the source code, I mean, you don't 334 00:19:20,280 --> 00:19:20,960 have access. 335 00:19:21,100 --> 00:19:22,960 Nikolay: Many companies don't understand that. 336 00:19:22,960 --> 00:19:26,060 I first felt this. 337 00:19:26,840 --> 00:19:31,020 I was experiencing this insight, what you're just sharing, that 338 00:19:31,020 --> 00:19:34,580 source code is like… it's like kind of deeper documentation. 339 00:19:36,100 --> 00:19:41,760 When I first had it in 2018, helping some company which we are 340 00:19:41,760 --> 00:19:45,320 migrating to RDS, they were huge already. 341 00:19:45,560 --> 00:19:50,540 They needed logical replication to first to Vertica, then to 342 00:19:50,540 --> 00:19:51,800 Snowflake, doesn't matter. 343 00:19:52,200 --> 00:19:55,760 But they considered some tools and 1 of the tools was maybe 1 344 00:19:55,760 --> 00:19:59,020 of the popular at that time was Click, Xatunity. 345 00:20:00,660 --> 00:20:03,740 And the documentation didn't cover so many questions I had. 346 00:20:04,860 --> 00:20:09,520 And I just, like, I realized I was talking to my client, I just 347 00:20:09,520 --> 00:20:13,680 said, you know what, if they had a source available, not open 348 00:20:13,680 --> 00:20:16,700 source, a source available, I would just answer myself, because 349 00:20:16,700 --> 00:20:20,640 I can read code, it doesn't matter language, I could find this 350 00:20:20,640 --> 00:20:23,740 place, how exactly they create logical slot. 351 00:20:24,140 --> 00:20:27,900 Spent a lot of time and efforts, I realized they create logical 352 00:20:27,900 --> 00:20:31,080 slot using SQL function, not replication connection. 353 00:20:31,500 --> 00:20:32,940 That was their problem. 354 00:20:33,700 --> 00:20:39,520 But for me, lack of source code means lack of ultimate documentation, 355 00:20:39,680 --> 00:20:40,180 basically. 356 00:20:41,260 --> 00:20:43,580 Michael: Ultimate and accurate and up-to-date. 357 00:20:46,160 --> 00:20:47,760 Nikolay: It's also about transparency. 358 00:20:48,100 --> 00:20:51,960 If you don't provide source code, you always can say, oh, we 359 00:20:51,960 --> 00:20:54,360 fixed this problem, but not in all cases. 360 00:20:54,680 --> 00:20:59,060 But you cannot bullshit if you have source available with all 361 00:20:59,060 --> 00:21:02,860 versions and tags and releases, like anyone can see. 362 00:21:02,860 --> 00:21:03,980 It's about transparency. 363 00:21:04,700 --> 00:21:09,060 And transparency is why I have doubts in Postgres. 364 00:21:09,840 --> 00:21:10,700 Michael: Oh, interesting. 365 00:21:11,180 --> 00:21:13,440 Nikolay: Yes, but it's not technical. 366 00:21:13,440 --> 00:21:15,660 Let's talk about technical first before we... 367 00:21:15,660 --> 00:21:18,420 Michael: Well, I think we can intertwine them. 368 00:21:18,420 --> 00:21:21,800 But a couple of other reasons I chose it that are on the non-technical 369 00:21:22,160 --> 00:21:22,640 side. 370 00:21:22,640 --> 00:21:25,520 I think when you're building for. 371 00:21:25,520 --> 00:21:29,240 So I I'd worked at places that chose Postgres, but I wasn't the 372 00:21:29,240 --> 00:21:30,000 decision maker. 373 00:21:30,000 --> 00:21:30,280 At those. 374 00:21:30,280 --> 00:21:33,580 You know, It was an already made decision when I worked there. 375 00:21:33,680 --> 00:21:36,980 I never chose explicitly until I started my own business. 376 00:21:37,300 --> 00:21:42,800 So it was only then that I mean, we chose it not just as users, 377 00:21:42,800 --> 00:21:46,620 of course, but as as a platform to build on. 378 00:21:47,080 --> 00:21:51,100 And because I was trying to build something like long term, it's 379 00:21:51,100 --> 00:21:55,620 very much, you know, a small, slow, steady, not VC backed, hopefully 380 00:21:55,640 --> 00:21:57,480 for decades, that kind of business. 381 00:21:58,180 --> 00:22:02,440 I didn't want to build for a platform that there was, that was 382 00:22:02,440 --> 00:22:06,020 like, that could shift and change drastically from underneath 383 00:22:06,020 --> 00:22:11,760 me and also that was really at risk of a company changing their 384 00:22:11,760 --> 00:22:13,520 mind or changing something drastically. 385 00:22:13,580 --> 00:22:17,040 So it's often described as platform risk and it felt really low 386 00:22:17,040 --> 00:22:20,060 with Postgres, not just because it was built on really solid 387 00:22:20,060 --> 00:22:25,120 foundations but also because there's no 1 company that can, if 388 00:22:25,120 --> 00:22:28,520 they, you know, there's no company behind it, there's no commercial 389 00:22:28,680 --> 00:22:32,100 entity that is dependent on its success. 390 00:22:32,100 --> 00:22:35,340 Nikolay: Like MySQL AB, right? 391 00:22:36,300 --> 00:22:37,280 Michael: Yeah, well, but... 392 00:22:37,280 --> 00:22:38,600 Nikolay: Or MongoDB, like the company. 393 00:22:38,600 --> 00:22:40,020 Michael: Yeah, like all of... 394 00:22:40,020 --> 00:22:42,240 If you consider all of the companies that have changed their 395 00:22:42,240 --> 00:22:46,220 licenses in the last few years, for example, That's just one example 396 00:22:46,220 --> 00:22:48,660 of how things can shift from underneath you. 397 00:22:48,900 --> 00:22:52,700 But also, they could go completely closed source, or in MySQL's 398 00:22:52,820 --> 00:22:57,940 case, stop publishing test cases, which is one step towards, again, 399 00:22:58,080 --> 00:23:01,400 I don't think we've had any awful examples where businesses really 400 00:23:01,400 --> 00:23:04,920 would have suffered, but they can if there's a single company 401 00:23:04,920 --> 00:23:05,800 behind it. 402 00:23:05,800 --> 00:23:09,220 And the fact that Postgres was not just, not just had seemingly 403 00:23:09,380 --> 00:23:13,260 had policies in place to prevent that happening, but also seem 404 00:23:13,260 --> 00:23:14,560 to live those values. 405 00:23:15,060 --> 00:23:20,040 There's a real, there's lots of companies that employ contributors 406 00:23:20,500 --> 00:23:24,400 and there's real collaboration between those on new features 407 00:23:24,400 --> 00:23:25,640 and on bug fixes. 408 00:23:25,680 --> 00:23:30,800 And it seemed truly collaborative, truly a community project, 409 00:23:30,860 --> 00:23:34,160 even though lots of those community members were, almost all 410 00:23:34,160 --> 00:23:38,600 were being paid to work on it, it was still a community, a group 411 00:23:38,600 --> 00:23:42,720 effort between collaborative organizations that could all benefit 412 00:23:42,720 --> 00:23:43,800 from it getting better. 413 00:23:44,060 --> 00:23:48,260 And it just didn't seem like it was as likely to have a fundamental 414 00:23:48,420 --> 00:23:50,900 shift as some of the other platforms out there. 415 00:23:51,100 --> 00:23:52,620 Nikolay: It's interesting, actually. 416 00:23:53,300 --> 00:23:58,940 It resonates with my idea that sometimes VC-backed company… Recently, 417 00:23:58,980 --> 00:24:03,180 we discussed companies which are closing, post-VC-backed companies, 418 00:24:03,180 --> 00:24:04,300 startups, right? 419 00:24:05,140 --> 00:24:09,960 It resonates with the idea that if you… VC-backed, you have pressure, 420 00:24:09,960 --> 00:24:10,340 right? 421 00:24:10,340 --> 00:24:13,880 And you can have… because of that pressure, you can make some 422 00:24:14,380 --> 00:24:18,180 unpredicted moves, unpredictable for your users and customers, 423 00:24:18,260 --> 00:24:18,760 right? 424 00:24:19,020 --> 00:24:23,360 For example, if you're a great successful Postgres startup, but 425 00:24:23,360 --> 00:24:28,240 not successful enough for VCs, you suddenly can decide to switch 426 00:24:28,240 --> 00:24:32,800 license and make more money if you lose some users, right? 427 00:24:33,180 --> 00:24:33,680 Yeah. 428 00:24:34,240 --> 00:24:37,740 This is about Postgres, but you apply the same logic for... 429 00:24:38,300 --> 00:24:41,940 This is about Postgres startup, Postgres-related startup, right? 430 00:24:41,940 --> 00:24:45,460 But you apply the same logic to Postgres core. 431 00:24:46,220 --> 00:24:48,500 Michael: Yeah, as a platform to build on. 432 00:24:49,400 --> 00:24:50,520 Nikolay: Yeah, this is interesting. 433 00:24:51,220 --> 00:24:56,580 And if, for example, distributed nature of development basically, 434 00:24:56,580 --> 00:24:57,080 right? 435 00:24:57,180 --> 00:24:58,760 Distributed nature of development. 436 00:24:58,820 --> 00:25:04,760 But of course there are some key people and if they have gone, 437 00:25:04,760 --> 00:25:09,560 for example, disappeared or decided to quit or something, Postgres 438 00:25:09,920 --> 00:25:12,280 has more chances to survive, right? 439 00:25:12,880 --> 00:25:17,300 This is what Bruce Momjian describes in his talk Will Postgres 440 00:25:17,300 --> 00:25:19,480 Live Forever, right? 441 00:25:19,480 --> 00:25:22,580 Because like Postgres was dead several times already. 442 00:25:23,300 --> 00:25:24,740 And someone picked it. 443 00:25:25,160 --> 00:25:29,180 Well, it was dead after it was basically closed in Berkeley. 444 00:25:30,020 --> 00:25:35,560 And then it resurrected in the middle of 90s as open source, 445 00:25:36,140 --> 00:25:36,640 right? 446 00:25:37,620 --> 00:25:39,180 So, yeah. 447 00:25:41,000 --> 00:25:41,940 Michael: Yeah, you're right. 448 00:25:41,940 --> 00:25:46,880 There are a few individuals who contribute a huge, huge, huge 449 00:25:46,880 --> 00:25:47,380 amount. 450 00:25:48,700 --> 00:25:53,740 But it also feels like there are others who contribute hugely 451 00:25:53,760 --> 00:25:54,160 too. 452 00:25:54,160 --> 00:25:59,580 And obviously, if a few of them or several of them stopped, like 453 00:25:59,580 --> 00:26:06,420 when MySQL joined Oracle via Sun, and you had the split with 454 00:26:06,420 --> 00:26:09,680 MariaDB if there was a fork like that in the Postgres world for 455 00:26:09,680 --> 00:26:13,140 some reason and several of the contributors all left at once 456 00:26:13,140 --> 00:26:17,960 or 1 or 2 really really big important ones did I can imagine 457 00:26:17,960 --> 00:26:25,240 it being detrimental to the speed of Postgres enhancements, but 458 00:26:25,240 --> 00:26:28,440 I do think there's still so much momentum. 459 00:26:28,680 --> 00:26:30,480 And actually, I don't think... 460 00:26:30,480 --> 00:26:34,200 I mean, we can talk about it, but I think feature growth is really 461 00:26:34,200 --> 00:26:38,180 helping with Postgres, but I actually don't think it's as, now 462 00:26:38,180 --> 00:26:41,600 that it's competitive with Oracle and SQL Server on a bunch of 463 00:26:41,600 --> 00:26:46,640 like enterprise and like performance features, it's not as important 464 00:26:46,640 --> 00:26:49,620 that we keep innovating as quickly on the feature level. 465 00:26:49,700 --> 00:26:54,220 Perhaps things like pgvector and making sure we're available 466 00:26:54,220 --> 00:26:58,640 for new use cases via extensions or via core could be important. 467 00:26:59,280 --> 00:27:01,160 But yeah, I'd be keen to go back to you. 468 00:27:01,160 --> 00:27:02,200 You already raised... 469 00:27:02,960 --> 00:27:06,220 Do you want to cover technical reasons why others are choosing 470 00:27:06,220 --> 00:27:09,560 it, maybe us or others, or do you want to move to doubts and 471 00:27:09,560 --> 00:27:09,840 things 472 00:27:09,840 --> 00:27:10,240 Nikolay: like that? 473 00:27:10,240 --> 00:27:11,960 Let's quickly cover technical reasons. 474 00:27:11,960 --> 00:27:18,460 So first of all, it has a strong foundation, follows principles, 475 00:27:18,640 --> 00:27:23,400 follows standard, not fully, but quite well, much better than 476 00:27:23,400 --> 00:27:24,180 many others. 477 00:27:24,920 --> 00:27:30,700 And this is like, it has principles like extensibility and so 478 00:27:30,700 --> 00:27:30,900 on. 479 00:27:30,900 --> 00:27:36,300 Like these technical things, like Extensibility is constantly 480 00:27:36,300 --> 00:27:38,240 bringing new features. 481 00:27:39,960 --> 00:27:47,240 pgvector relies on extensions framework, and it has its own lifecycle, 482 00:27:47,440 --> 00:27:50,180 but maybe it will go to the core at some point. 483 00:27:50,840 --> 00:27:51,720 Maybe, right? 484 00:27:51,980 --> 00:27:56,080 As the full-text search or XML did in the past, they were separate 485 00:27:56,840 --> 00:28:00,080 extensions, but then went to core. 486 00:28:00,580 --> 00:28:03,740 And this means also features are coming, coming, coming. 487 00:28:03,740 --> 00:28:04,500 That's great. 488 00:28:05,820 --> 00:28:09,720 Michael: Or it could, there's another success case where it could 489 00:28:09,720 --> 00:28:12,980 be more like PostGIS, which has been incredibly successful and 490 00:28:12,980 --> 00:28:16,360 helpful for the growth of Postgres in my opinion, and has stayed 491 00:28:16,360 --> 00:28:17,060 an extension. 492 00:28:17,300 --> 00:28:21,540 And I think a lot of people are choosing PostGIS, and Postgres 493 00:28:21,600 --> 00:28:24,480 comes along for the ride, rather than the other way around. 494 00:28:24,480 --> 00:28:25,400 And that's really cool. 495 00:28:25,400 --> 00:28:27,180 And it could be the same in pgvector. 496 00:28:27,400 --> 00:28:30,480 Like, people could be choosing pgvector, and Postgres comes along 497 00:28:30,480 --> 00:28:31,900 for the ride in those cases. 498 00:28:32,120 --> 00:28:36,100 So yeah, either case could be a huge success still. 499 00:28:37,060 --> 00:28:39,020 Nikolay: PostGIS is an interesting example. 500 00:28:39,860 --> 00:28:43,980 Yeah, and it's kind of a whole different world, right? 501 00:28:44,640 --> 00:28:49,280 But imagine it was part of, like Postgres documentation doesn't 502 00:28:49,280 --> 00:28:50,040 have PostGIS. 503 00:28:51,460 --> 00:28:58,700 Would Postgres be even more successful if PostGIS was in core 504 00:28:58,700 --> 00:29:00,300 and documentation had it? 505 00:29:00,300 --> 00:29:07,540 Or, for example, if Timescale, the company, decided to go all 506 00:29:07,540 --> 00:29:14,440 in with TimescaleDB and make it super open, eventually contributing 507 00:29:14,540 --> 00:29:15,420 it to Core. 508 00:29:16,240 --> 00:29:19,540 Just imagine, with all compression and all the stuff and Postgres 509 00:29:19,540 --> 00:29:20,880 documentation had it. 510 00:29:21,740 --> 00:29:25,520 I know it's not good for their business maybe, but imagine this 511 00:29:25,520 --> 00:29:26,980 world we live in. 512 00:29:27,160 --> 00:29:33,220 Postgres has PostGIS, it has TimescaleDB, pgvector, it has everything, 513 00:29:33,240 --> 00:29:33,740 right? 514 00:29:33,960 --> 00:29:38,460 The level of quality is unified, right? 515 00:29:38,460 --> 00:29:43,240 It's like it's being tested all the time in the same manner as 516 00:29:43,440 --> 00:29:44,440 Postgres itself. 517 00:29:44,660 --> 00:29:49,780 And people don't spend time trying to understand how to bring 518 00:29:49,780 --> 00:29:50,920 these pieces together. 519 00:29:51,400 --> 00:29:55,080 And if it's managed Postgres like RDS or something, all the pieces 520 00:29:55,080 --> 00:29:56,100 are there already. 521 00:29:56,940 --> 00:29:58,780 Michael: Yeah, I can see arguments both ways. 522 00:29:58,780 --> 00:30:02,040 I've seen people I respect a lot arguing completely the opposite 523 00:30:02,040 --> 00:30:02,540 direction. 524 00:30:02,640 --> 00:30:04,780 Nikolay: So maybe we should define what Postgres is. 525 00:30:04,780 --> 00:30:09,780 There is Postgres when you just install it with a few extensions, 526 00:30:09,780 --> 00:30:13,320 not a few, dozens of them, so-called KDRIP modules, right? 527 00:30:13,320 --> 00:30:18,025 And also CLI, tool psql, and that's it, right? 528 00:30:18,025 --> 00:30:18,580 That's it. 529 00:30:19,900 --> 00:30:25,180 It also can create a user in your Ubuntu or Debian or CentOS 530 00:30:25,180 --> 00:30:28,940 and set things up and be installed as a service. 531 00:30:30,100 --> 00:30:30,740 Of course. 532 00:30:30,920 --> 00:30:31,920 But that's it. 533 00:30:32,580 --> 00:30:37,540 There are also super popular extensions which make Postgres very 534 00:30:37,540 --> 00:30:42,840 different, like PostGIS or TimescaleDB or pgvector. 535 00:30:43,860 --> 00:30:44,360 Great. 536 00:30:44,540 --> 00:30:46,020 It's also Postgres or no? 537 00:30:46,020 --> 00:30:47,280 It's also Postgres, right? 538 00:30:47,280 --> 00:30:48,220 Bigger Postgres. 539 00:30:50,280 --> 00:30:55,260 It's like there is Los Angeles, but if you live in Long Beach, 540 00:30:55,260 --> 00:30:56,980 you still live in Los Angeles, right? 541 00:30:56,980 --> 00:31:02,120 It's like there are bigger Postgres. 542 00:31:02,520 --> 00:31:05,340 Michael: Al-Khalili That's a great example because it depends 543 00:31:05,340 --> 00:31:08,520 in what context you're speaking and who you're speaking to and 544 00:31:08,520 --> 00:31:11,180 you know how familiar they are with the area if they say where 545 00:31:11,180 --> 00:31:15,140 are you from but it's if you if you're a local coffee shop in 546 00:31:15,660 --> 00:31:18,620 near LA and someone asks where you're from, you'll probably be 547 00:31:18,620 --> 00:31:19,300 more specific. 548 00:31:20,940 --> 00:31:24,140 But if you're in Thailand and somebody asks where you're from, 549 00:31:24,140 --> 00:31:25,580 you'll probably just say LA. 550 00:31:25,680 --> 00:31:26,440 Nikolay: Same here. 551 00:31:26,760 --> 00:31:28,040 Michael: Same, exactly the same. 552 00:31:28,040 --> 00:31:31,300 Nikolay: We use Postgres, but if you say we use Timescale Cloud, 553 00:31:31,300 --> 00:31:34,320 we're already like, okay, also Postgres, but definitely with 554 00:31:34,320 --> 00:31:38,240 TimescaleDB and probably without some things they don't provide. 555 00:31:40,080 --> 00:31:40,520 Interesting. 556 00:31:40,520 --> 00:31:41,900 So there's bigger Postgres. 557 00:31:42,180 --> 00:31:45,800 There are also many tools around, like you're building 1 and 558 00:31:45,800 --> 00:31:46,860 we're building something. 559 00:31:47,160 --> 00:31:52,080 So these tools, yeah, but a lot of options and it's like quite 560 00:31:52,080 --> 00:31:53,060 a rich ecosystem. 561 00:31:53,460 --> 00:31:54,360 Great right? 562 00:31:54,720 --> 00:31:57,380 So this is also a reason why Postgres, because the ecosystem is 563 00:31:57,380 --> 00:31:59,640 super rich and growing and it's great. 564 00:32:00,400 --> 00:32:03,160 Michael: Although possibly, like I think that could be an argument 565 00:32:03,160 --> 00:32:04,000 against it. 566 00:32:04,000 --> 00:32:07,040 Definitely when I first was looking at Postgres, I'd actually 567 00:32:07,040 --> 00:32:10,740 say that other ecosystems were a little bit more advanced. 568 00:32:10,880 --> 00:32:15,560 There were more people building external tools for some of the 569 00:32:15,560 --> 00:32:17,060 other systems. 570 00:32:17,320 --> 00:32:22,120 So I do think there's like an, that's not as obvious a 1 for 571 00:32:22,120 --> 00:32:25,260 choosing Postgres in my opinion, but it's great that the more 572 00:32:25,260 --> 00:32:30,360 and more are popping up the last like technical ones I had were obviously 573 00:32:30,360 --> 00:32:33,840 the extensibility, not just in terms of pre-made extensions, 574 00:32:33,840 --> 00:32:37,520 but the fact you could like as a user, you can extend Postgres 575 00:32:37,660 --> 00:32:38,400 in various ways. 576 00:32:38,400 --> 00:32:41,520 I think a lot of people do choose it for those reasons. 577 00:32:41,720 --> 00:32:42,260 Nikolay: Of course. 578 00:32:42,260 --> 00:32:42,760 Yeah. 579 00:32:42,800 --> 00:32:48,160 Starting from the creation of your own data type to your own type 580 00:32:48,160 --> 00:32:54,220 of index and extensions and now like even more things up to storage 581 00:32:54,220 --> 00:32:55,420 engines and so on. 582 00:32:55,760 --> 00:32:57,080 Michael: Yeah, exactly. 583 00:32:57,340 --> 00:33:00,140 I think people can dip their toe in with a new data type and 584 00:33:00,140 --> 00:33:02,640 then like gradually get more and more excited by what they could 585 00:33:02,640 --> 00:33:02,880 build. 586 00:33:02,880 --> 00:33:05,860 And then the last technical 1 I had on my list was, and this 587 00:33:05,860 --> 00:33:08,860 might be more why people stay rather than why people join in 588 00:33:08,860 --> 00:33:12,440 the first place, but backwards compatibility is so helpful. 589 00:33:13,320 --> 00:33:17,480 Because of the good fundamentals and because Postgres has five-year 590 00:33:18,080 --> 00:33:23,680 support of major versions, people aren't forced to upgrade like 591 00:33:23,680 --> 00:33:26,520 that often and things don't tend to break when they do. 592 00:33:26,520 --> 00:33:29,060 You know, obviously we've done episodes on this and there's a 593 00:33:29,060 --> 00:33:31,680 lot of complexity around major version upgrades and you do have 594 00:33:31,680 --> 00:33:33,220 to re-release those things. 595 00:33:34,000 --> 00:33:39,620 But compared to other systems, compared to other platforms that 596 00:33:39,620 --> 00:33:42,840 people are building and building for, things like there are major 597 00:33:43,100 --> 00:33:48,060 breaking changes in a lot of database major version upgrades. 598 00:33:48,580 --> 00:33:51,180 And having seen people have to deal with those and have to get 599 00:33:51,180 --> 00:33:55,520 stuck on really old versions, it's just so much better in the 600 00:33:55,520 --> 00:33:56,540 Postgres world. 601 00:33:56,580 --> 00:34:00,280 So I think that might be a reason people stay or like maybe some 602 00:34:00,280 --> 00:34:02,320 very smart people, the reason they choose it. 603 00:34:02,320 --> 00:34:05,060 But I definitely wasn't aware of that when I was first coming 604 00:34:05,060 --> 00:34:05,820 across it. 605 00:34:07,840 --> 00:34:08,340 Nikolay: Right. 606 00:34:08,420 --> 00:34:08,920 Yeah. 607 00:34:09,800 --> 00:34:10,080 Michael: Yeah. 608 00:34:10,080 --> 00:34:14,680 So on the doubt side of things, you mentioned transparency. 609 00:34:15,140 --> 00:34:15,640 Transparency. 610 00:34:16,020 --> 00:34:16,360 Nikolay: Yeah. 611 00:34:16,360 --> 00:34:24,620 For me, transparency is a key factor I dislike in Postgres because 612 00:34:25,580 --> 00:34:28,040 the source code is open. 613 00:34:28,500 --> 00:34:30,500 In mailing list, they are open as well. 614 00:34:30,500 --> 00:34:31,420 Tanner Iskra Sure. 615 00:34:31,420 --> 00:34:36,760 Leon Mandeley But if you look closer, and I know like people 616 00:34:36,760 --> 00:34:39,300 who just use Postgres, they don't see problems. 617 00:34:40,160 --> 00:34:44,240 But I know for sure, not like only based on my experience, but 618 00:34:44,240 --> 00:34:48,580 based on many companies, not just 1, not just 2, many companies, 619 00:34:49,300 --> 00:34:53,160 people who are in the head of those companies, Postgres-related 620 00:34:53,500 --> 00:34:59,240 companies, big ones, big names, they feel this lack of transparency 621 00:34:59,680 --> 00:35:00,520 in processes. 622 00:35:03,480 --> 00:35:08,440 Like Postgres project doesn't apply this principle to itself. 623 00:35:10,080 --> 00:35:11,600 Processes are not transparent. 624 00:35:12,870 --> 00:35:18,380 Michael: Question, just to ask where you're going with this, 625 00:35:19,900 --> 00:35:23,600 do you consider any other alternative projects more transparent 626 00:35:23,720 --> 00:35:26,540 and more open, like, along this axis? 627 00:35:27,980 --> 00:35:31,060 Nikolay: I'm honestly not a big expert here. 628 00:35:31,060 --> 00:35:38,140 I just see if, like, well, usually organizations which are built 629 00:35:38,400 --> 00:35:43,020 recently, which are quite strong and so on, have kind of democracy, 630 00:35:43,200 --> 00:35:43,700 right? 631 00:35:44,060 --> 00:35:47,520 They have elections, some rotation of power. 632 00:35:47,900 --> 00:35:50,740 Michael: When I'm looking at alternatives, if I'm thinking like 633 00:35:50,740 --> 00:35:57,100 MySQL, Redis, SQLite, none of those have as far as... 634 00:35:57,180 --> 00:36:00,000 Nikolay: But it doesn't mean there's no problem. 635 00:36:00,820 --> 00:36:01,920 Michael: No, no, no. 636 00:36:02,100 --> 00:36:07,320 I'm just like the question why Postgres partly asks if not Postgres, 637 00:36:07,440 --> 00:36:08,160 what else? 638 00:36:08,180 --> 00:36:09,640 Like, they're almost in the database. 639 00:36:11,140 --> 00:36:12,120 Or do you know what I mean? 640 00:36:12,120 --> 00:36:16,320 Like if you're gonna, if you doubt it and then say, maybe this 641 00:36:16,320 --> 00:36:18,340 isn't it, what would it be? 642 00:36:18,340 --> 00:36:19,740 Like what's the alternative? 643 00:36:21,040 --> 00:36:22,160 Nikolay: Yeah, good question. 644 00:36:22,540 --> 00:36:26,380 It can be forked postgres with different principles. 645 00:36:26,540 --> 00:36:27,040 Interesting. 646 00:36:27,620 --> 00:36:28,300 Yeah, yeah. 647 00:36:28,620 --> 00:36:33,420 Honestly, I like, I, When I raised in the beginning of this discussion, 648 00:36:33,580 --> 00:36:37,460 I raised this, I have doubts, I have them for long. 649 00:36:38,800 --> 00:36:42,160 I just wanted to look inside myself and validate. 650 00:36:42,440 --> 00:36:45,860 But while we are discussing all these things are good, I'm still 651 00:36:45,860 --> 00:36:46,740 staying honestly. 652 00:36:46,740 --> 00:36:50,700 Like I already feel like I'm still staying with Postgres, but 653 00:36:50,720 --> 00:36:54,020 people must know I'm not fully satisfied with things. 654 00:36:54,840 --> 00:36:59,120 And at some point, I might say it's enough for me. 655 00:37:00,040 --> 00:37:01,280 I don't feel it right now. 656 00:37:01,280 --> 00:37:02,580 I'm very far from it. 657 00:37:03,040 --> 00:37:05,380 But lack of transparency. 658 00:37:06,160 --> 00:37:10,380 Michael: Yeah, so we've got lots of reasons why we chose Postgres 659 00:37:10,380 --> 00:37:14,340 in the first place, lots of positives, like still that we would 660 00:37:14,340 --> 00:37:17,200 choose it again today and that we do continue to choose it every 661 00:37:17,200 --> 00:37:17,700 day. 662 00:37:17,800 --> 00:37:19,780 Some areas for improvement for sure. 663 00:37:19,780 --> 00:37:21,920 I don't see a great alternative. 664 00:37:22,040 --> 00:37:25,920 I think it's still exceptional and love building for it, and 665 00:37:25,920 --> 00:37:28,100 I'm hoping to for decades to come. 666 00:37:29,600 --> 00:37:33,520 It'd be cool to hear from anybody that works in or with like 667 00:37:33,520 --> 00:37:37,660 similar organizations with different governance structures and 668 00:37:37,660 --> 00:37:38,800 how those work. 669 00:37:38,960 --> 00:37:42,040 All I can think of is like benevolent dictator types you know 670 00:37:42,040 --> 00:37:45,620 Linux and and SQLite and things but I'm sure there are other 671 00:37:45,620 --> 00:37:49,040 good communities like I think the Kubernetes 1 keeps coming up 672 00:37:49,040 --> 00:37:49,900 as a... 673 00:37:49,900 --> 00:37:51,240 Nikolay: Cloud native, maybe. 674 00:37:51,820 --> 00:37:53,920 Michael: Yeah, so there are some interesting ones out there that 675 00:37:53,920 --> 00:37:56,040 are run slightly differently, so it'd be great to hear from people 676 00:37:56,040 --> 00:37:57,600 that are heavily involved in those. 677 00:37:58,780 --> 00:37:59,280 Nikolay: Mm-hmm. 678 00:38:00,140 --> 00:38:00,560 Right. 679 00:38:00,560 --> 00:38:06,780 So, yeah, in my opinion, Postgres is definitely one of the greatest 680 00:38:06,780 --> 00:38:08,840 examples of open source. 681 00:38:11,580 --> 00:38:18,980 I see it as a big bazaar with one big cathedral in the center, 682 00:38:19,940 --> 00:38:25,140 which now I recognize and smaller cathedrals on this bazaar maybe. 683 00:38:25,320 --> 00:38:29,240 Well, the cathedral is one, it's one, but it's still a bazaar and 684 00:38:29,240 --> 00:38:33,100 no matter what, I think it has good chances to be a good bazaar, 685 00:38:33,420 --> 00:38:38,220 like with many voices, even if people inside the cathedral in the 686 00:38:38,220 --> 00:38:40,740 center don't agree with those voices, right? 687 00:38:41,320 --> 00:38:47,180 These voices cannot be due to probably due to original nature, 688 00:38:47,800 --> 00:38:52,260 license chosen in the '90s, and the original nature of the project 689 00:38:52,260 --> 00:38:57,840 itself, even if there are conflicts, this bazaar will still be 690 00:38:57,840 --> 00:38:58,340 huge. 691 00:38:58,660 --> 00:39:03,500 And I think it has good chances to be huge, with different opinions, 692 00:39:03,760 --> 00:39:05,040 sometimes very contradicting. 693 00:39:06,140 --> 00:39:11,400 So yeah, I'm glad it's a bazaar because I think it's a good thing, 694 00:39:11,400 --> 00:39:12,560 like an open source. 695 00:39:12,560 --> 00:39:15,740 You know, I'm referring to cathedral and the bazaar by recreation. 696 00:39:16,560 --> 00:39:19,080 Michael: Yeah, yeah, I love the analogy and you've used it multiple 697 00:39:19,080 --> 00:39:22,440 times in the past but it's always been cathedral or bazaar. 698 00:39:22,720 --> 00:39:23,160 Nikolay: LM. 699 00:39:23,160 --> 00:39:26,990 Yeah, but now I see bazaar with a big cathedral in the center. 700 00:39:26,990 --> 00:39:27,040 AD. 701 00:39:27,040 --> 00:39:28,420 Michael: There is a cathedral in the middle. 702 00:39:28,420 --> 00:39:30,080 Maybe it's not that big though. 703 00:39:30,140 --> 00:39:34,020 It is a cathedral but maybe the bazaar is bigger and the cathedral 704 00:39:34,020 --> 00:39:34,448 is very large. 705 00:39:34,448 --> 00:39:34,572 LM. 706 00:39:34,572 --> 00:39:34,696 I 707 00:39:34,696 --> 00:39:37,260 Nikolay: feel sometimes this cathedral in the center tries to 708 00:39:37,260 --> 00:39:39,060 pretend there is no bazaar almost. 709 00:39:40,960 --> 00:39:43,320 Or this cathedral can define all the rules. 710 00:39:43,320 --> 00:39:44,020 It cannot. 711 00:39:44,240 --> 00:39:46,060 It cannot define all the rules. 712 00:39:46,560 --> 00:39:47,640 Michael: Have you seen Monty Python? 713 00:39:47,640 --> 00:39:50,820 There's Monty Python, I think it's, which 1 is it, maybe Search 714 00:39:50,820 --> 00:39:51,860 for the Holy Grail? 715 00:39:52,360 --> 00:39:53,560 Have you seen the Monty Python 716 00:39:53,560 --> 00:39:53,940 films? 717 00:39:53,940 --> 00:39:54,020 Nikolay: No. 718 00:39:54,020 --> 00:39:56,020 Michael: No, I think you would like them. 719 00:39:56,720 --> 00:39:58,060 Nikolay: My son loves it. 720 00:39:58,380 --> 00:39:58,780 He tries 721 00:39:58,780 --> 00:40:01,080 to convince me for years to watch everything. 722 00:40:01,080 --> 00:40:01,400 Michael: You have 723 00:40:01,400 --> 00:40:03,220 to watch it, I think you'll enjoy it. 724 00:40:03,340 --> 00:40:07,820 But there's 1 moment where the king is going up to the peasants 725 00:40:07,820 --> 00:40:11,960 working the land and tells them what to do because he's their 726 00:40:11,960 --> 00:40:12,260 king. 727 00:40:12,260 --> 00:40:14,520 And they're like, I don't believe in a king. 728 00:40:15,060 --> 00:40:16,220 So it's quite funny. 729 00:40:16,500 --> 00:40:17,560 I'll link that up. 730 00:40:17,560 --> 00:40:19,340 Nikolay: Yeah, I should see. 731 00:40:19,840 --> 00:40:25,480 So yeah, for sure it will remain to be a cathedral and it's good. 732 00:40:26,060 --> 00:40:34,120 So yeah, I thank you for help validating that I still want to 733 00:40:34,120 --> 00:40:35,580 use Postgres, you know. 734 00:40:35,660 --> 00:40:38,980 For me, I have enough reasons to use Postgres no matter what. 735 00:40:39,060 --> 00:40:40,220 No matter what happened. 736 00:40:40,440 --> 00:40:41,020 I mean... 737 00:40:41,040 --> 00:40:41,540 Michael: Nice. 738 00:40:41,540 --> 00:40:44,060 Maybe we'll discuss again in a couple of years and see where 739 00:40:44,060 --> 00:40:44,880 you are there. 740 00:40:45,020 --> 00:40:46,020 Nikolay: Yeah, well, yeah. 741 00:40:46,020 --> 00:40:50,820 I'm quite positive I will keep using it and help other people 742 00:40:51,000 --> 00:40:52,740 to use it in a better way. 743 00:40:53,860 --> 00:40:54,640 Michael: Me too. 744 00:40:55,440 --> 00:40:56,960 Wonderful, thanks so much, Nikolay. 745 00:40:56,980 --> 00:40:57,940 Catch you next week.