Monday, August 27, 2007

DataObjects.Net 4.0 Platform Vision

DataObjects.Net 4.0 Platform Vision can be found here. Some sections aren't written yet.

Posted by Alex Yakunin

11 comments:

  1. Pavel ŠavaraAug 27, 2007 07:19 AM
    Go, Xtensive, go! :-)
    ReplyDelete
  2. Yves HanoulleAug 27, 2007 09:35 AM
    Nice to see this published.
    Thank you for mentioning me.
    Glad to help you out.
    I was surprised to see that you used Team City.
    ReplyDelete
  3. AnonymousAug 27, 2007 10:40 AM
    Nice to see some news out of X-tensive, unfortunately I am afraid you guys are falling into a trap. You are trying to accomplish too much, DO.NET 4.0 is going to be vapour for a long time, and by the time you release, the world will have moved on.
    ReplyDelete
  4. Alex YakuninAug 27, 2007 11:35 AM
    We're targeting to release alpha \ CTP in the near two months. Certainly only key features will work; probably the only storage type it will support will be Memory storage - it is relatively simple (no schema upgrade is required), perfect for testing everything + debugging query tranformation engine for non-SQL storages. The next step will be alpha with SQL storage support.

    So we want to deliver it ASAP.

    P.S. The world always moves forward - that's why we've changed some of our plans (in particular, decided to support LINQ even in CTP). We anyway should provide something more attractive then ADO.Net Entity Framework - many features I listed here will be supported by it as well, but in general we want to win the comparison.
    ReplyDelete
  5. Alex YakuninAug 28, 2007 12:15 PM
    Still writing... Today and yesterday I'v got big discussions with developers related to v4.0. Should log the conclusions in this document as well.
    ReplyDelete
  6. Alex YakuninAug 29, 2007 01:39 PM
    I just updated the document. Still unfinished, but brief summaries are provided for almost all sections that aren't written yet.
    ReplyDelete
  7. Alex KofmanAug 30, 2007 11:54 AM
    It's very interesting to see complete version of document (-:
    ReplyDelete
  8. Alex YakuninSep 9, 2007 10:20 AM
    The document is updated once more. Almost finished.
    ReplyDelete
  9. AnonymousSep 14, 2007 03:25 PM
    Really, why, why, why? Why are you guys trying to be everything to everyone? Your OR/M is brilliant, you guys should focus on bringing that up to .NET 3.0 technologies such as LINQ and lightweight SQL engines such as SQLite and the embedded SQL 2005 version.

    Instead, you guys are reinventing the wheel here, trying to build a persistence engine that be decoupled from an RDBMS. This is going to be an engineering nightmare, why don't you stand on the shoulders of giants and capitalize on the RDMBS, instead of reinventing B+ Tree's and indexes?

    DO.NET 4 has been on the drawing board for two years now, and I fail to see any real incentive to even check it out, since it looks like it's going to be a huge bloated mess. It looks like a death march, I think you guys should step back and re-evaluate what the customers really want and not build things for the sake of engineering itself!
    ReplyDelete
  10. Alex YakuninSep 20, 2007 11:54 AM
    > Why are you guys trying to be everything to everyone?

    May be because we feel we can ;)

    > You guys should focus on bringing that up to .NET 3.0 technologies such as LINQ and lightweight SQL engines such as SQLite and the embedded SQL 2005 version.

    As you see, we're bringing more then support of such engines. Implementing support for the listed ones will be relatively simple task after all.

    Btw, we agree LINQ support should be provided in any case now.

    > Instead, you guys are reinventing the wheel here, trying to build a persistence engine that be decoupled from an RDBMS.

    That's not quite true:
    - We build such engine because we see a good way to build it.
    - We anyway need it - think about index range caching (quite good feature) and offline storages (must-have feature).
    - Finally (here you'll be completely disappointed), we're working on clustered index implementation. Since indexing support is the key requirement for any DO storage now, it will be relatively easy to get DO running on cluster.

    > This is going to be an engineering nightmare

    Hopefully not.

    > Why don't you stand on the shoulders of giants and capitalize on the RDMBS, instead of reinventing B+ Tree's and indexes?

    We don't reinvent anything: B+ tree is relatively simple and well-known data structure. Really, its non-concurrent implementation is comparable to e.g. Hashtable by its complexity. We simply use it, as well as many other similar classes, that aren't built into .Net (e.g. Set, PriorityQueue and so on).

    > DO.NET 4 has been on the drawing board for two years now, and I fail to see any real incentive to even check it out, since it looks like it's going to be a huge bloated mess.

    Spending ~ one year (that's what we really have) on development of such a project is actually not too much.

    > It looks like a death march, I think you guys should step back and re-evaluate what the customers really want and not build things for the sake of engineering itself!

    I agree we're probably trying to deliver too much. It is really not quite interesting to reinvent the wheel, but if you decided to do this, the only way to win is to invent (and create) a Superwheel ;)
    ReplyDelete
  11. AnonymousSep 26, 2007 02:12 PM
    We don't want SuperWheels. We need your Titanium Hubcaps 4.0 to cover up our beat-up wheels ASAP~

    go...xtensive..go...
    ReplyDelete

Subscribe to: Post Comments (Atom)