Tuesday, October 15, 2002

Clueless Clueless. A few things that have been bothering me about .NET recently: A 20MB runtime requirement for client side .NET WinForms apps Around $6000 upfront software costs for ASP.NET deployment (XP Server plus SQL Server 2000). I really like the XML libraries (fast, stable, great error reporting), but the lack of source code for the libraries is a huge problem. There is no production runtime for an OS other than Windows, and there doesn't appear to be one coming anytime soon. Mono is great and I think will be a force to be reckoned with, but not for a while. [...] [Ken Rawlings] US$6000 startup costs for developing? And they wonder why linux, perl, php, python and java ran away with the developers? Honestly, is there anyone home in Redmond? [Brett Morgan's Insanity Weblog Zilla] I wonder how many windows-only developers left the platform over the last 5 years? Its very hard to measure; typically BigCo's do it by downloads which isn't a good number. I remember in the old days when you used to find teams of VB programmers at companies writing rich front ends to databases or Office macros that only worked on windows. I just don't see those kinds of people any more. Maybe its just the circles I move in but most people seem to build web apps these days - and when it comes to web apps people typically don't want to filter out part of the user base, be they Mac, Linux, Unix or phone users. Its probably gonna take some pretty clever use of their monopoly power (like funky Office & IE gizmos that only work from inside .NET) for MS to win back anything like the only-develop-for-Windows developers they used to have. 6:47:28 PM    comment []

Database and code round tripping continued Me and my big mouth :-) Synergy. This summer, Matthias Bohlen, Ara and I had some discussions about Integration of UML2EJB, Middlegen and XDoclet. Mathias brought up the issue yesterday, Ara was quick to respond, and James has caught up. Many smart people with the same objective: integrate more tools. Database Metamodel: I took a closer look at the commons-sql API and I must say the model package is quite similar to Middlegen's API for the same things (Table,Column,Relationship,RelationshipRole). The APIs were built for different purposes, and I think they have a lot to learn from each other. I would really like to have only one API for this. Well, James already suggested we work together, and I'm all for it. Groovy! commons-sql's XML model is very database-centric and doesn't have room to specify notions like cardinality and directionality of relationships. I believe this information would have to be included in both the XML and the Java model, at least if Middlegen is to use it. If we want to generate XMI from database models, we need this too. No problem, we can do that. Code metamodel: James talks about output of XJavadoc, and I'm not quite sure what he means. XJavadoc (just like QDox) is just a java parser that builds an object model of a java source. No output, except that XJavadoc let's you write the source back to a writer after you have modified it via its API. In this scenario, you'd only want to modify (add) some @tags and then write it back. Sorry my mistake. For some reason I thought XJavadoc could output its model as XML. Maybe we just need a single API to the XJavadoc/QDox API and that can suffice as the code metamodel? We could always come up with a standard XML representation of the model if thats helpful. I noticed from the XDoclet 2 design notes, that XJavadoc and QDox could be merging? If so maybe the common API could be the code metamodel? But what kind of classes are we talking about here? EJBs? JDOs? Other classes? If we want to have an ubiqutous code metamodel based on @tags, we must standardise the database-related @tags beyond the way XDoclet has done it today. XDoclet has @ejb.persistence column-name="foo". Likewise "proprietary" tags for JDO. Perhaps we need something as simple as @sql.mapping column-name="foo" instead? Then we could deprecate these ejb and jdo tags and have a standard set of @tags for java <-> sql mapping. Think about it. Conclusion: The whole idea here is deciding on some common tools and formats. I'm still a bit confused about what belongs where, but I'll get over it :-) [Aslak Helles[macron]y's Weblog] Me too :-) 1:09:09 PM    comment []

New Java Blog, Aslak Via an email conversation with Aslak I've spotted Aslak Helles[macron]y's Weblog. Well worth reading on all things XDoclet and middlegen. You can subscribe to it via RSS here. 12:04:12 PM    comment []

Database and code round tripping Java Persistence Frameworks, again. [Memory Dump] Ara has some interesting thoughts on database model <-> code <-> UML round tripping.  middlegen also looks interesting. We should try to define a metamodel for the database, for the code and for the UML then we could plug in different tools for moving from one model to the other and performing diffs etc. Then a whole bunch of code and tools could just work at the database metamodel layer. Or the code metamodel layer etc. For UML its looking like XMI is the official OMG answer; though it is kinda verbose; I'd be happy with a cut down less verbose syntax specialised just for Class models. This is metamodel where things like Together/J work. For the code metamodel, I guess we could always just use Java code itself with doclet tags? Or failing that maybe the output of, say, XDoclet's Xjavadoc? We've already tools like XDoclet and QDox that work fine at this metamodel. For a database metamodel, I was hoping the commons-sql's model (either the beans or the XML format) could be used as the metamodel. Its really just an attempt to define a logical database model on which code could operate. middlegen has a very similar model - I wonder if we could work together and share a common relational model? Then the relational metamodel could be specified via XML, can be auto-created from a database connection or can be created from UML or the code metamodel via middlegen / XDoclet etc. If nothing else, if middlegen did share (or adapt to) the commons-sql model then middlegen would get SQL generation for different databases for free. 10:58:55 AM    comment []

Microsoft Puts SourceForge Clone Into Beta Microsoft Puts SourceForge Clone Into Beta [Slashdot] Interesting. I think MS has its work cut out for itself to build a large open source community anywhere near the sizes of Linux, GNU or Jakarta. Especially if its under a dodgy MS licence. All I've seen so far have been C# ports of existing Java open source projects (like NAnt, NUnit etc). Is there really much of a C# open source community out there? 9:51:58 AM    comment []