Announcement

Collapse
No announcement yet.

PostgreSQL 11 Beta 2 Released With VACUUM & XML Fixes

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • PostgreSQL 11 Beta 2 Released With VACUUM & XML Fixes

    Phoronix: PostgreSQL 11 Beta 2 Released With VACUUM & XML Fixes

    One month has passed since PostgreSQL 11 Beta 1 while today the second beta has succeeded it...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    My biggest gripes with PostgreSQL is the use of a non-standard license, which although it is free open source software it does contribute to license proliferation. Also that on Debian/Ubuntu it clutters the file system with lots of binaries in /bin/. Not all prefixed with pg.

    Other than that, it is a nice database. But so is MySQL (since version 8), and Microsoft SQL Server (supports data masking which is good for GDPR and security, also supports stored procedures written in C#).
    PostgreSQL have support for JSON, although I don't think it have support for the SQL/JSON standard which Oracle Database does.
    Also if you want to update a JSON document you can't update only the affected property, you must update the whole document. I think MongoDB supports more flexible updates which is useful on larger documents.

    I wonder what the future of databases is.
    Will be all be locked into proprietary cloud databases such as Azure Cosmos DB that are offered as SaaS / DaaS, or will self-hosted open source databases still be a thing.
    Will we be stuck with the SQL language forever or will there be some new, modern database language? SQL is from 1974.

    Comment


    • #3
      I'd love to see the new publication/subscription model of logical replication actually replicate DDLs. Not sure how this was released minus that, to be honest. Otherwise, loving the improvements.

      Originally posted by uid313 View Post
      Will we be stuck with the SQL language forever or will there be some new, modern database language? SQL is from 1974.
      Ah, it's not the latest thing, so it's no longer relevant? I've heard that one before. I was asked @ work a couple of months back why the data warehousing team are using some kind of NoSQL technology for everything that we do. I answered: "because we're not fucking idiots". The relational model, and SQL for querying it, has been around for so long precisely because it's the best way of handling the data needs of most projects. The strange edge-cases where some other model is allegedly better - key/value stores etc - only apply in the most extreme of cases, and are used successfully by about 2 or 3 companies worldwide. The flock that follows them because they read a blog post about it from Facebook or whatever - come and ask them a couple of years down the track how things are going for them. I've seen plenty of companies go down the hadoop/hive path in this fashion ... but I haven't seen a single one do it *successfully* ... because it's just an insane model for regular data warehousing workloads.

      Comment


      • #4
        Originally posted by dkasak View Post
        I'd love to see the new publication/subscription model of logical replication actually replicate DDLs. Not sure how this was released minus that, to be honest. Otherwise, loving the improvements.



        Ah, it's not the latest thing, so it's no longer relevant? I've heard that one before. I was asked @ work a couple of months back why the data warehousing team are using some kind of NoSQL technology for everything that we do. I answered: "because we're not fucking idiots". The relational model, and SQL for querying it, has been around for so long precisely because it's the best way of handling the data needs of most projects. The strange edge-cases where some other model is allegedly better - key/value stores etc - only apply in the most extreme of cases, and are used successfully by about 2 or 3 companies worldwide. The flock that follows them because they read a blog post about it from Facebook or whatever - come and ask them a couple of years down the track how things are going for them. I've seen plenty of companies go down the hadoop/hive path in this fashion ... but I haven't seen a single one do it *successfully* ... because it's just an insane model for regular data warehousing workloads.
        I agree that NoSQL was overly hyped and is not the solution to everything. I agree that relational databases still have a place, and that relational databases are generally preferred over NoSQL solutions except in edge cases.

        Programming languages have seen much improvements and new generations with new features, concepts and syntax. However, query languages have not, its basically been just SQL and iterations of it.
        Today nobody uses COBOL, B, ALGOL, Fortran, etc. It would be a pain in the ass to use such old, outdated, legacy languages when there are much nicer modern languages.

        Perhaps there could be a query language that is better than SQL. That is easier to understand and reason about. That is more portable and better standardized.
        There is D4 / Dataphor, but it seems it have not gained much traction.

        Maybe SQL is ugly, clumsy, confusing, tricky and intuitive.
        Maybe it would be nice to have functionality like lambdas, regular expressions, etc. Maybe something like LINQ (from .NET Framework) or ReQL (from RethinkDB). Maybe something like XQuery or XPath. Maybe something like CSS/DOM query selectors.

        Comment


        • #5
          Originally posted by uid313 View Post
          Today nobody uses COBOL, B, ALGOL, Fortran, etc. It would be a pain in the ass to use such old, outdated, legacy languages when there are much nicer modern languages.
          The latest COBOL standard is COBOL 2014, and it's still used quite a lot, especially in big organisations (banks, governments, multinationals, etc.).
          The latest FORTRAN standard is Fortran 2018, and it's still used a lot for scientific computation stuff (e.g. on supercomputers).

          It's not because they aren't hipster languages that they are outdated or aren't used much.

          Originally posted by uid313 View Post
          Maybe it would be nice to have functionality like lambdas, regular expressions, etc.
          PostgreSQL & many other SQL implementations already support regular expressions.

          Comment


          • #6
            Originally posted by uid313 View Post
            My biggest gripes with PostgreSQL is the use of a non-standard license, which although it is free open source software it does contribute to license proliferation. Also that on Debian/Ubuntu it clutters the file system with lots of binaries in /bin/. Not all prefixed with pg.


            Will we be stuck with the SQL language forever or will there be some new, modern database language? SQL is from 1974.
            What is a standard license ? Something that adheres to a restrictive gpl-style? It's basically a Bsd license so is no more or less hard to adhere to than a bsd license. Or am I missing something? To me bsd licenses are the best short of public domain which is true freedom.

            As to sql being old, i'm not sure of the issue because old does not necessarily correlate to disfunctional, surely. Pick-like databases gave been around since the 60s. Still functional, still relevant. I guess sql language is so widely well known it would take a long time to usurp it - that and the biggest commercial provider is oracle.

            Comment


            • #7
              Originally posted by JanC View Post
              The latest COBOL standard is COBOL 2014, and it's still used quite a lot, especially in big organisations (banks, governments, multinationals, etc.).
              The latest FORTRAN standard is Fortran 2018, and it's still used a lot for scientific computation stuff (e.g. on supercomputers).

              It's not because they aren't hipster languages that they are outdated or aren't used much.
              Yes, the standards for old languages like COBOL, Fortran, etc are still updated.
              Yes, these languages are still used in some big organisations. However it is to maintain a old legacy code base. Nobody writes any new software in these languages. No young people learn these languages.
              These are not pleasant languages to work with.

              Originally posted by Bsdisbetter View Post
              What is a standard license ? Something that adheres to a restrictive gpl-style? It's basically a Bsd license so is no more or less hard to adhere to than a bsd license. Or am I missing something? To me bsd licenses are the best short of public domain which is true freedom.
              No, I don't prefer the GPL over the BSD.
              A standard license is one that is approved by the FSF and OSI and DFSG.
              The BSD license is a good license, I like the BSD license.
              Yes, the PostgreSQL License is basically a BSD license, but the problem is that it is not the BSD license.
              So it contributes to license proliferation.

              Originally posted by Bsdisbetter
              As to sql being old, i'm not sure of the issue because old does not necessarily correlate to disfunctional, surely. Pick-like databases gave been around since the 60s. Still functional, still relevant. I guess sql language is so widely well known it would take a long time to usurp it - that and the biggest commercial provider is oracle.
              Yeah while SQL is old it is not dysfunctional, but perhaps complex, cumbersome and inconvenient.
              This is the first time I have heard about Pick.

              Comment


              • #8
                Originally posted by uid313 View Post
                No, I don't prefer the GPL over the BSD.
                A standard license is one that is approved by the FSF and OSI and DFSG.
                The BSD license is a good license, I like the BSD license.
                Yes, the PostgreSQL License is basically a BSD license, but the problem is that it is not the BSD license.
                So it contributes to license proliferation.
                I guess then it's a sign of the eco-system that is FOSS (most commonly GNU/Linux). So many distributions, so many applications. Personally, I don't like any of these organisations that seek to impose restraints on licenses and tell developers what they should do. If you want to issue a license asking people to pat your dog in return for use, then so be it, as it's your software.

                However, we wouldn't have an issue if everyone released software under BSD-style or PD licenses. Then you don't have to read reams of pages of agreement to attempt to understand what your rights are. You don't need 20000 words to tell you how to use those BSD/MIT but you sure do for GPL style licenses:



                (That's insane!)

                But, I have to agree the license proliferation can be a cause for concern if you are GPL and your intention is to basically hijack anything that comes within your purview. For most users, a license is something they click before installing the software. Most licenses are basically unenforceable, some break local laws and most are so damn wordy you just ignore them. Most people who use a GPL-style license have NO idea what it encompasses, how it can (or cannot) be enforced and just how basically communistic it is.

                Originally posted by uid313 View Post
                Yeah while SQL is old it is not dysfunctional, but perhaps complex, cumbersome and inconvenient.
                This is the first time I have heard about Pick.
                Perhaps there is a correlation between the simplicity of database system and the language required to extract information from it?
                Perhaps complexity is because, while there's a standard, very few adhere fully to it and have their own add-ons to the language. Oracle is different to MariaDB/MySQL/RDB/DB2/Sqlite etc etc.. If you stick to one or maybe two of these, it's not too hard, it's when you have to use 3 or more it does become complex.

                Comment


                • #9
                  Originally posted by Bsdisbetter View Post
                  If you want to issue a license asking people to pat your dog in return for use, then so be it, as it's your software.
                  Then it is not free software.

                  Originally posted by Bsdisbetter View Post
                  However, we wouldn't have an issue if everyone released software under BSD-style or PD licenses. Then you don't have to read reams of pages of agreement to attempt to understand what your rights are. You don't need 20000 words to tell you how to use those BSD/MIT but you sure do for GPL style licenses:
                  Yes, there would still be a problem, since if it is a BSD-style license, but it is not the BSD license, then it is a vanity license that contributes to license proliferation.

                  Originally posted by Bsdisbetter View Post
                  But, I have to agree the license proliferation can be a cause for concern if you are GPL and your intention is to basically hijack anything that comes within your purview. For most users, a license is something they click before installing the software. Most licenses are basically unenforceable, some break local laws and most are so damn wordy you just ignore them. Most people who use a GPL-style license have NO idea what it encompasses, how it can (or cannot) be enforced and just how basically communistic it is.
                  License proliferation is a concern regardless of what license you prefer, it has nothing to do with GPL per se.
                  Perhaps most users can click pass the license before reading it, but in the corporate world it is different, then you can't just ignore licenses and everything has to pass through company lawyers, and that cost money.

                  Originally posted by Bsdisbetter View Post
                  Perhaps there is a correlation between the simplicity of database system and the language required to extract information from it?
                  Perhaps complexity is because, while there's a standard, very few adhere fully to it and have their own add-ons to the language. Oracle is different to MariaDB/MySQL/RDB/DB2/Sqlite etc etc.. If you stick to one or maybe two of these, it's not too hard, it's when you have to use 3 or more it does become complex.
                  The standard is rather poorly defined and overly verbose. Such as CHARACTER VARYING(n).
                  Maybe it would be nicer if create table accepted a annotated data structure or the query consisted of chained functions.

                  Perhaps the ReQL and LINQ approach is easier?
                  Perhaps the XPath or XQuery approach to querying is easier?

                  I don't know. Maybe SQL is a good language. I am just not convinced. I imagine there could be something better.

                  Comment


                  • #10
                    Originally posted by uid313 View Post
                    Then it is not free software.


                    Yes, there would still be a problem, since if it is a BSD-style license, but it is not the BSD license, then it is a vanity license that contributes to license proliferation.


                    License proliferation is a concern regardless of what license you prefer, it has nothing to do with GPL per se.
                    Perhaps most users can click pass the license before reading it, but in the corporate world it is different, then you can't just ignore licenses and everything has to pass through company lawyers, and that cost money.


                    The standard is rather poorly defined and overly verbose. Such as CHARACTER VARYING(n).
                    Maybe it would be nicer if create table accepted a annotated data structure or the query consisted of chained functions.

                    Perhaps the ReQL and LINQ approach is easier?
                    Perhaps the XPath or XQuery approach to querying is easier?

                    I don't know. Maybe SQL is a good language. I am just not convinced. I imagine there could be something better.
                    Have you gone and asked the PostgreSQL peeps what the exact reason for them not using a common license is? Maybe there no longer isn't any and they could switch

                    Comment

                    Working...
                    X