Comments on: In Praise of Draconian Error Handling, Part 2

By: John Cowan

John Cowan — Wed, 28 Oct 2009 15:22:36 +0000

Actually, that isn’t even a legal relative reference, never mind a legal absolute URL: “%SV” isn’t decodable as a %-escape.

By: Simone

Simone — Mon, 29 Jun 2009 11:05:48 +0000

This is a good approach.

By: Albert Nonymouse

Albert Nonymouse — Wed, 17 Jun 2009 00:35:27 +0000

You’re all idiots. Use XOM to develop your internal XML apps. Sleep easy knowing that any XML generated by your internal apps will be processed by all of the more permissive parsers used by your partners. Cry a little bit because XML generated by your external partners may trigger errors on import. Get over it when your XOM driven preprocessing XML cleanup tool helps give you clean data for imports into your internal database without much overhead. Laugh to the bank as your competitors are dragged down by the inefficiencies in their systems, while your own processes run smooth as butter.

By: C. Doley

C. Doley — Thu, 11 Jun 2009 21:08:16 +0000

Your conclusion is wrong. Whereas it might be nice to have a mode which rejects all but perfect XML, saying that XOM won’t change when XML is less than perfect is basically telling your users to switch to other parsers than XOM. Imagine what would happen if I released a web browser that rejected everything except perfect HTML. It would not be able to render 90% of the web. Do you think anyone would want to use such a browser, even if it was “correct”?

I happen to really like XOM, but if I encountered a case such as this I’d simply switch to a different parser.

By: Jason Osgood

Jason Osgood — Sat, 06 Jun 2009 06:34:36 +0000

This isn’t a code bug. It’s a spec bug. XML entities are evil. A good rule of thumb: anything tough to describe is probably wrong.

By: Joshua Haberman

Joshua Haberman — Fri, 05 Jun 2009 18:10:14 +0000

You should check out this thought experiment for a counterpoint to this attitude. Summarized: yes, it helps you find bugs, but it also DoS's you when these bugs surface (instead of gracefully degrading), and the bugs might not even be your bugs.

By: Mike

Mike — Fri, 05 Jun 2009 16:03:31 +0000

On the one hand, I completely agree with your approach here.

On the other hand, this is just one more thing that makes me want to run screaming from XML. This stuff is way more complicated than any Real Person(TM) is going to be able to deal with.

By: Michael Kay

Michael Kay — Fri, 05 Jun 2009 15:12:32 +0000

One of the nice things about writing open source software is that you can tell your users they are wrong. That’s one of the reasons open source software is often higher quality. However, sometimes I think you have to give up the good fight, because sticking to your principles is just making your users angry. I did that for example when I started supporting the Sun “jar:” URI scheme (which doesn’t follow the rules for URI schemes); I’m also slowly giving up on the issue of accepting Windows filenames where URIs are expected (partly because most of the library routines I rely on seem to allow them). So I think you’re right in theory, but sometimes practical compromises are needed.

By: Norman Walsh

Norman Walsh — Fri, 05 Jun 2009 14:31:14 +0000

Nice catch. Bravo to XOM!