Re: PingTool.py converts 8 bit characters into entities
Georg Bauer <gb-BRhJDZTO+/[email protected]> Wed, 17 Mar 2004 09:33:36 +0100
| Newsgroups | gmane.comp.pythin.pyds.devel |
|---|---|
| Message-ID | <r02010100-1028-C9112F1F77ED11D8A517000A9573A72A@[10.0.0.145]> |
Hi! > PingTool.py converts 8 bit characters to &#xxx; entities and send ping > to local community server . But if document encoding of PyCS was > UTF-8, web browsers fail to display UTF-8 entities. Yeah, pinging is actually really disgusting. I used to send out stuff in my encoding, but that did break. So I changed everything to entities, but that's no solution as you point out. Ok, what I just commited to CVS is a way to specify the behaviour on server level. You can choose wether HTML entity encoded strings should be sent, or wether utf-8 encoded strings should be sent. Default is entity encoding, so you need to go to your preferences for pinging and set those servers that are utf-8 capable to that setting. Default for new installations is utf-8 capability for technorati and blo.gs, entity encoding for weblogs.com and your local community server (as that might be a Radio Community Server that behaves much like weblogs.com with respect to charsets). Please try it out. I didn't do too extensive testing, so things might not work as expected. The main problem with pinging is weblogs.com, actually. A rather shitty ping server, as it doesn't implement utf-8 correctly, neither iso-8859-1, nor entities :-/ It just stores what it get's on byte level and returns that in a HTML page without charset declaration, so that is assumed as ISO-8859-1. So utf-8 pings are broken -you get sent out utf-8 strings, but mixed with ISO-8859-1 strings and without declaration of charset. If you send in ISO-8859-1 strings, it mostly works. Except when weblogs.com has some problem - it still returns mac-roman encoded strings sometimes. And sometimes ISO-8859-1 strings are in the XMLRPC result. But no charset declaration in the XMLRPC result. So expart barfs on parsing the XML. It's the old story - Userland tools are broken with regard to charset handling. I am thinking wether it's really worth to support such a lousy implementation or wether it's not better to just throw that one out :-/ bye, Georg