22 mei 2004

Word is out: ObjectSpaces to be merged with WinFS

Exciting news: ObjectSpaces will be merged with WinFS!

The URL that I was giving was slightly off: go to http://msdn.microsoft.com/data/ to read all about.

Although there were many guesses on what the news would be nobody really hit the spot.

My first impressions on this are twofold:

  1. The API will be tightly integrated into Longhorn and therefore become an integral part of the operating system. This makes it even more important in my view than the previous scenario of an “add-on” to ADO.NET.
  2. It's a delay for even longer than “shortly after Whidbey“.

From my email exchanges with Andrew Conrad I was giving this extra information as well:

“Basically, we are merging ObjectSpaces and WinFS into an unified O/R mapping API and framework for Longhorn.  This means the OS API will change significantly and that it will be delayed until Longhorn.  However, most of the POCO scenarios are deemed valuable and will be supported.“

To read a bit more on his views on POCO (Plain Old CLR Objects) go read this and this on his weblog.

So the API will change a lot. I fear that there will be no release of the code as it stands today. This will probably be good news (is it?) to the builders of O/R mapping technologies, because the time-gap between now and Longhorn will have to be bridged.

My questions to you all are: what do you think of this announcement? I know there must be people out there that have very explicit ideas and feelings on this.

One important question I have myself: how does the WinFS picture fit with SQL Server 2005? I feel that ObjectSpaces as an O/R mapper has a certain place for developers when it comes to generalizing your DAL to a relational datastore. Although WinFS is moving in that direction, it still seems that SQL Server serves a slightly/hugely different purpose.

Also, keep your eyes on Andrew's weblog, as he will be posting additional information on ObjectSpaces and WinFS (probably today as well. But I may have lost a little credibility on release times of news ;-) ).

Comments are welcome. Fire ahead.

Comments

# Frans Bouma said:

This must be a major blow to the guys working on Objectspaces, but then again, as I said earlier, there must have been something going on behind the scenes for a long time, as they didn't had a lot of progress looking at the builds that were released. This is of course good news for O/R mapper vendors, at least I see it as good news. Your analysis about WinFS serving a different purpose than SqlServer is IMHO correct. What I wonder though is: MBF is based on Objectspaces as well. Does this mean that objectspaces is simply a 'generic' technology which is integrated in WinFS and MBF or is MBF using a different object layer?

zaterdag 22 mei 2004 9:37
# Paul Gielens said:

Just great, knowing I need to saddle my customers with expensive Longhorn licenses (and hardware). I'd rather see ObjectSpaces live on top of v1.* incarnations of the framework. This makes me to seriously consider Paul W's mapper which is as good as ObjectSpaces API compliant and available today. I just wished MS acknowledged we need ORM today to compete with the far superior platform (ORM wise) Java. My Java coworkers are laughing their brains out if I tell them I still need to wait a few more years in order to practice ORM out of the box. Not that there aren't enough alternatives ;)

zaterdag 22 mei 2004 11:11
# Frans Bouma said:

Paul: what are they using then? Hibernate? That's a non-standard api as well as the api's currently available on .NET. Java has one big advantage over .NET: it has a framework in place which allows you to plug in O/R mapping logic and keep a fully distributed environment. However I think for example hibernate is not fully utilizing it (unless it's used in an appserver like JBoss). People today who are waiting for Objectspaces are not busy with real life development with real deadlines and real customers. MS isn't the company to provide you with all the controls and tools to do your job, it's the company to provide definitions and frameworks IN which you should do your job. Despite the critizism poured over J2EE and EJB, it is at least a standard a lot of tools commit to, something Sun has done right and MS simply doesn't understand. However that doesn't mean you can't get your job done today. There are plenty of O/R mapping technologies available today for .NET, which are used in real life projects which can be sometimes very large (how does 1100 tables and 60 developers sound ? :D). That has puzzled me for some time: people get angry because Objectspaces gets delayed or canned. Why? There is plenty of  stuff available TODAY which does even more than what objectspaces was able to do over a year. I hope the flip-flopping of the focus around Objectspaces will proof that a big company behind a tool isn't a guarantee for support in the long run. (if at all, how many fixes are there released for .NET 1.1?)

zaterdag 22 mei 2004 12:10
# Paul Gielens said:

OJB> Despite the critizism poured over J2EE and EJB, it is at least a standard a lot of tools commit to, something Sun has done right and MS simply doesn't understandIsn't this because of Sun actually giving priority to things that matter to it's developers instead of their product's market opportunities.

zaterdag 22 mei 2004 12:35
# Frans Bouma said:

"Isn't this because of Sun actually giving priority to things that matter to it's developers instead of their product's market opportunities."Could be. It can also be because Sun has failed to integrate something to make money of Java. I mean: Microsoft's equivalent of EJB is COM+. This is tied to the OS. Clever move. COM+ is great, the integration with .NET however is poorly done (should be transparent, not by inheriting from a base class, distributed apps have to use multiple techniques instead of relying on 1 layer) but might get better with longhorn. (I'm pretty sure it will).What's indeed weird is why MS doesn't do the same as sun did NOW: the technology to base it on is already established and available. (and why doesn't this blogengine remember my settings when I check 'Remember Me' :D)

zaterdag 22 mei 2004 13:26
# Alex Thissen said:

Wouldn't it be a great idea/gesture of Microsoft to post the ObjectSpaces source code as it stands today into a GotDotNet workspace? The community could maintain/bugfix this code. The ObjectSpaces API is going to "change significantly" as Andrew states it. So, putting this up on GDN right now will give developers something to work on. It will get a different twist and API when integrated into WinFS anyway. When Whidbey ships, so could an O/R mapping implementation that holds the nice ideas that ObjectSpaces has (Generics, non-intrusive model, mapping files, et cetera).Then again, this might just be wishful thinking.

zaterdag 22 mei 2004 13:37
# Paul Gielens said:

Great idea Alex, but I don't think MS would be willing to publish x years of work for free? And if they would, in what what way could MS benefit from such a move? Besides most projects on GDN tend to fail anyway.

zaterdag 22 mei 2004 13:57
# Alex Thissen said:

Browsing around a little I found out that Carlo Pinasco of Microsoft had already spilled the beans on ObjecSpaces and WinFS yesterday, just before he went on vacation to Corsica:http://www.ugidotnet.org/PermaLink.aspx?guid=84014396-d41c-4948-89db-c767136764e0"Microsoft is merging the development efforts behind a feature "ObjectSpaces" into related work on the Windows "Longhorn" object data system code-named "WinFS".This decision reduces the confusion that the overlapping feature sets of ObjectSpaces and WinFS would cause. Both technologies aim to reduce the barrier between application data and application objects. This decision allows developers to focus on one set of APIs for that functionality."

zaterdag 22 mei 2004 14:18
# Alex Thissen said:

Paul, good questions. They would maybe gain some credit for making such a move. My impression on ObjectSpaces as it stands is that it is pretty much feature complete. That makes it different from other projects that start from scratch. I agree that the percentage of failure might be high on those projects.

zaterdag 22 mei 2004 14:23
# Mats Helander said:

WinFS + ObjectSpaces...doesn't that equal Cairo/OFS?

zaterdag 22 mei 2004 14:25
# Paul Wilson said:

I think its sad that MS can't get an O/R Mapper out with all their resources, but then again it also looks like their attitude has been more of a case of throwing resources at it instead of getting the proper and real experienced resources.  Of course, this also does prove that people should be more than willing to trust the current 3rd party O/R Mapper vendors, at least as much as they can trust Microsoft anyhow.  :)  The biggest problem from my perspective isn't that there are good solutions out there, but that most Microsoft shops simply won't consider anything that isn't from Microsoft, unless its another silly grid.

zaterdag 22 mei 2004 15:43
# Theo Albers said:

The migration from ObjectSpaces to WinFS means that Microsoft only has to support MSSqlServer in my opinion. WinFS already uses the MSSql-engine. You must have a very good reason using Oracle for instance on a Windows machine when MSSqlServer 2005:- is able to store .Net objects - supports stored procedures written in .Net- is tightly coupled with the .Net object-relational persistence frameworkMSSqlServer with .Net support, ORM-api (WinFS) and Visual Studio.Net integration looks like Oracle with Java support and JDeveloper package to me.... I don't have a problem with that, but some ISV's have to support multiple database platforms.-Theo

zaterdag 22 mei 2004 16:32
# Matt said:

This is terrible news. Here are my reasons (outside of the obvious delay):1) I just want ObjectSpaces to do "1 thing good" . . . and that's bridge the object/relational gap to make it easier for my apps to persist data. Period! There are enough problems in that arena, alone, to solve . . . without the hassle of integration into WinFS and the problem IT creates. 2) As already mentioned, this surely means ObjectSpaces is only interested in MSSQLServer . . . believe it or not, many of my customers are not interested in either licesning MSSQL or utilizing MSDN . ..  many of my customers prefer a "lite" version of my app running against Access . . . and i prefer cashing checks so I deliver to them.3) I think MS might want to re-familiarize themselves w/ YAGNI and KISS. . .perhaps, even, tuck their ego a bit under their belts and just ship a solution to solve a problem without trying to solve all problems (including many they have created on their own). The PDC bits were actually working quite well for alpha code . ..  then, the CTP build was released and functionality reverted. Now, WinFS goo is added and delays are into the longhorn timeframe. In this respect, have they ever heard of iterative development?Anyhow, I'm sure things will be GREAT when they're out . . . but, I guess i'm the ONLY developer having to actually release products for profit; so I turn to the vendors (and, in fact, had made an O/R mapper purchase just last week) . . . I guess MS thinks everyone else out there can write longhorn code and not have to worry about shipping something for 2+ years?

zondag 23 mei 2004 5:49
# Paul Gielens said:

@Theo:Which ORM did you purchase, and why?

zondag 23 mei 2004 13:27
# Jason said:

Is it just me or did WinFS in fact slip from Longhorn?http://yahoo.businessweek.com/magazine/content/04_16/b3879009_mz001.htmSo wouldn't a WinFS/OS rollup be post longhorn??Jason

zondag 23 mei 2004 20:31
# Matt said:

I purchased Objectz.NET (http://www.mongoosesolutions.com); because, outside of Paul Wilson's ObjectSpaces, it seemed to be the only actual "true" O/R mapper and not simply an "O/R mapper code generator" of sorts.Although, unfortunately I'm quite disappointed w/ the support I've received . . . so, ultimately, I'm still in the market.

maandag 24 mei 2004 14:50
# Scott said:

Matt,It is my understanding that WinFS will be included in Longhorn but that its functionality will be restricted to desktop usage and not available for use with network drives and such.Microsoft has really messed this O/R mapping thing up. It's even worse than it first seems too when you really sit down and think about the fact that we're still 2+ years away from getting our hands on a released product and actually being able to deploy apps to people, well, that is another year or so after that. After all, even if Longhorn is an overwhelming success you have to figure it to be at least a full year after RTM - MINIMUM - before you can start counting on any significant number of people to have it on their desktop. It's just insane! I wish they had thought this whole thing out a little better before revealing any of the preview versions of it. This is a technology that is needed now, not 3 or so years from now and it is my opinion that Microsoft has really dropped the ball on this one.-Scott

maandag 24 mei 2004 20:41