Simon Fell > Its just code > February 2005

PocketXMLRPC v1.2.1 released, updated to latest PocketHTTP, and Jim T Row's patch to vastly increase base64 decoding performance, thanks Jim!

Just a reminder that mainstream support for the MS STK ends at the end of March, get those migration to PocketSOAP projects moving!.

Saw Amon Tobin last night at Bimbo's, all the previous times I've seen him live its always been a great show, but last night was a real disappointment not a lot of recognizable Tobin tunes, no visuals at all, the whole night would of been a complete washout if it wasn't for Telephone Jesus Jim

Just released PocketHTTP v1.2.3, just a minor update, includes the ability to have timeouts longer than 32 seconds, a new SaveAs method to write the HTTP response body direct to disk (useful for handling images from eVB which doesn't like arrays of bytes), and improved error reporting for SSL over Proxy server problems.

If you're running Axis 1.1 (and why wouldn't you be, 1.2 is determined to stay in RC status), you really need to be aware of this (serious IMO) bug with its mustUnderstand processing. If you need a fix and want to stay on the 1.1 code base, you just need to incorporate this diff.

Regardless of what anyone (Vendor's in particular) will tell you, getting real world interop is hard (i.e. across more than 2 toolkits), Nelson will talking about this at ETech. Dare argues that doing contract first with the tools is better than doing it by hand (i.e. write the WSDL/XSD by hand) because if you do it by hand you'll use constructs that the tools don't like. Well, if you use the tools, then all that means is that you'll use constructs that one particular tool is ok with, but you'll have no idea how many other tools will / won't like it. If you're trying to build a service with wide ranging interop (like Nelson & I am) then you're pretty much stuck with having to try every little change with every toolkit you remotely care about, you can read as many WS-I docs as you like, they won't tell you that you really want to avoid trying to send a nil xsd:int to .NET.