[forty]

I meant for implementing the protocol of course. If you don't have to implement it, then of course it does not matter as much if it's bad or not.

At the time I was looking into it closely, there were issues with XML, like with the starttls messing the original document, there were issues with XML validation (for xml namespaces I think?), and other minors arrangements with the XML spec such that using a normal pull parsing lib was not enough to solve all problems. But it was quite a long time ago (10+ years), possibly that have been solved for a while.

[forty]

in term of practical impact of using XML like this, at the time, the server has to practically parse everything that was in the content of messages not addressed to the server itself. I would imagine a reasonable protocol should separate addressing from the content in such a way that the router doesn't have to parse all the content to do its job