|

|
| 2003-06-30 |
|
|
Actually there is a problem with the charset stuff in the XML-RPC spec. Or not in the spec, but in the implementations of the company whose former CEO created the spec (was that diplomatic enough? ). The problem is in the fact that almost all Userland software abuses the spec in the way that they don't give correct charset encodings in XML-RPC stuff.
They send ISO-8859-1 chars without declaring the XML stuff as ISO-8859-1 - this doesn't validate (XML charset by default is UTF-8 and many 8bit chars are not valid prefix codes in UTF-8), so you have to patch headers and parsers to communicate with Userland software (RCS or Radio), as the results from RCS or the requests and results from Radio (I am on the receiving side of Radio if I work on the Python Community Server) won't be allowed by validating XML parsers.
I don't speak of the chars that are not in ISO-8859-1 that are sent out by Radio from time to time (for example the smart quotes in error messages when you access non-existant methods on a Mac Radio installation), as that would be annoying but unproblematic if the stuff was declared as ISO-8859-1 - the parser would take it, it would just look funny.
Why do I care about this? Usually I don't do comment stuff on this blog. So why on this one? Simple reason: it makes the life of programmers unnecessary hard. Usual toolkits to do XML parsing need correct and valid XML to work, otherwise they barf (many toolkits are built around the excellent and fast expat parser).
So it would be nice if Userland would live up to the specs their former CEO built and would lead a way of technical excellence in the implementation of those specs.
|
|
posted at 10:58:40
#
|
|
| |
| 2003-06-29 |
Bug in medusa creates some small problems in PyDS |
|
There is a bug in medusa with regard to the request unescaping. Medusa does the unescaping (turning elements with %XX format into the corresponding char) on the whole request, not only on the script path. This poses problems if your parameters use escaping to transport chars with special meaning - the escaping get's unescaped and so the parsing of URLs looses elements or shortens them. For example you can't pass an URL with a fragment part as a parameter, as the unescaping will make everything behind the # char as the fragment of the full URL instead as part of the query parameter.
I reported it twice to the medusa-dev mailing list, but no reaction so far. Same with somebody else who stumbled over the same problem. I think I will put in my own http_server.py into PyDS to circumvent this error, as the maintainer of medusa doesn't seem to be interested in fixing this annoying error. 
Update: download the following file and put it into the PYTHONLIB/medusa/ path, overwriting your http_server.py copy. This should fix the problem.
http_server.py: 25.00 KByte, last change: Sun, 29 Jun 2003 21:32:14 GMT, type: text/x-python click here to download file
|
|
posted at 23:15:44
#
|
|
| |
| 2003-06-25 |
|
|
This release is presumeably the last for docutils 0.2. The next release (0.6) will be based on docutils 0.3. This release fixes lot's of small nagging bugs. Nothing big and fancy, just typical gruntwork to be done from time to time. |
|
posted at 13:29:04
#
|
|
| |
| 2003-06-24 |
|
|
Tim gives a nice short (ok, short for him ) consensus on why to embark something like Pie - a new syndication format, with community driven development.
I will look at what develops over at Sam's Wiki (and - as my time and knowledge permits - will contribute) and will use the Python Desktop Server as a testbed. So let's see what comes out of this.
Of course, the Python Desktop Server will allways support RSS, as those are the most common formats out there for syndication.
|
|
posted at 13:13:04
#
|
|
| |
| 2003-06-19 |
Blog! bookmarklet for PyDS |
|
If you want a bookmarklet that allows you to post to your running instance of Python Desktop Server, just drag the following link to your favorites:
Blog!
If you are on a page that you want to blog, just press the Blog! favorite and a new window will open with a posting entry for the title and link of the current page. |
|
posted at 15:33:52
#
|
|
| |
| 2003-06-10 |
|
|
There is a bug in PyXML 0.8.1 that leads to memory corruption and other weird bugs (segfaults mostly) with very large SOAP requests. This can be solved by upgrading to PyXML 0.8.2. The title links to the relevant posting in the SOAPpy mailing list.
In the setup.py of 0.8.2 there is a small bug where an import of distutils.sysconfig is missing. Just add this import and the install should work fine.
|
|
posted at 17:35:28
#
|
|
| |
| 2003-06-03 |
|
|
There were some minor problems in the stray links resolving in 0.5.0, so I thought to put out a fix before announcing on freshmeat.
|
|
posted at 10:50:08
#
|
| |
|
|
The new beta includes many feature enhancements. Most noteably the patches to the restructured text engine and the WikiTool by Garth are now part of the distribution. AccessRestrictions are part of the distributions. Templates have been reworked to make use of h2 tags (former versions started with h3 tags). Many patches and internal reorderings happened. Stray links in restructured text are now resolved against tools that support this resolving (currently the WeblogTool, the StoryTool and the WikiTool) - another patch by Garth.
Due to the massive amount of patches and changes, upgraders should check changes in the pyds.css file and the local templates - almost all of them changed in this release.
This release is somewhat experimental in parts, especially the new WikiTool.
If you installed some tools before in ~/.PyDS/tools/ (like the AccessRestrictionsTool or the WikiTool), please remove those and use the distributed versions.
As always, details are in the changelog.
|
|
posted at 10:07:28
#
|
|
|
This is the Python Desktop Server weblog.
 (Donations will be used by the author to buy stuff, fullfill selfish wishes or do other silly recreational things. You have been warned.).
The PyDS is
|