This week I muddled my way through the documentation for the new Barracuda 4.x API, and I was reminded of something I saw on Slashdot awhile back:

“XML is like violence: if it doesn't solve your problem, use more.”

Yes, instead of sending your query as key/value pairs the way a standard HTML form would work (which is how the Barracuda 3.x API worked), the new API takes queries in the form of XML data, much like Microsoft's autodiscover protocol which only ever passes a single value as an argument, but needs it to be wrapped in XML anyway.

This wouldn't annoy me so much, but Barracuda's documentation fails to show how to actually do a query using the Perl module that they themselves recommend. That's right, the documentation suggests using CPAN's XML::RPC, then never mentions anything about it again. Instead they give examples that use incomplete, truncated XML listings.

But at least I can get a list of domains, which I couldn't find a way to do with the 3.x API.

I've been doing some further experimentation with autodiscovery. First of all, while Apple supports autodiscovery for Exchange 2007 servers, they officially do not support it for POP3/IMAP servers. I've filed a bug report for a feature request, asking them to fix that.

I've also filed a bug report with Mozilla, but they seem opposed to the idea of adding support to Thunderbird. In some ways, I can understand why - they don't want Microsoft's autodiscovery protocol to become popular, because the protocol sucks; they want to design a better system and get everybody to support that instead. Of course this better system doesn't actually exist yet, and support for Microsoft's protocol could be added now, but if nobody wants to do it, it's not going to happen.

Microsoft has this great tool for troubleshooting autodiscovery issues: the Microsoft Exchange Server Remote Connectivity Analyzer. Unfortunately I've found not just one but four bugs in their autodiscovery implementation so far; clearly their test site doesn't share any actual code with Outlook, they just wrote a client implementation from scratch according to their understanding of the spec, and... they didn't do a very good job. I've reported the following bugs to Microsoft:

1. The requested URL is capitalized, /Autodiscover/Autodiscover.xml, which doesn't matter on IIS but breaks if you're using a server that enforces case sensitivity. Either the spec should say that these paths must be case-insensitive, or the implementation should match the spec and use lower-case paths.
2. Port numbers are ignored when following a redirect. If you redirect to https://example.com:8000/autodiscover/autodiscover.xml, it will try to connect to port 443 and fail.
3. XML redirects are not understood. The spec says you can either serve an XML file with <Action>redirectUrl</Action> or use a standard HTTP redirect, but the test site thinks getting an XML file with HTTP status 200 is an error.
4. According to the spec, redirects can only point to a URL that ends in “autodiscover.xml”; any other filename is not valid and Outlook wil reject it. The test site does not have this (very stupid) limitation.

I wonder if they'll bother to fix these...

You need to watch this. Keith Olbermann explains why we absolutely have to fix health care in this country, and what that really means. Please pass this around; people need to see it.

• Watch online
• 213 MB download
• Transcript

You can send me e-mail at contact@phroggy_NOSPAM.com or use this form: