of course, the user who reported the bug still thinks konqi is broken as they didn't think to switch user agents, so all isn't well. this is an all too common event and results in unwarranted marring of konqueror's reputation.
so half an hour of putzing about later and i have a command line KDE/Qt app (weighing in at ~70LOC) called "agents" that rips through a kio_httprc file and spits all the domain-name to user-agent mappings into a postgresql database. this simple little app is aided by a short shell script that does some batch processing of kio_httprc files. sweet.
now i can do things like:
agents=# select count(id), d.id, d.domain
from domains d join agents a on d.id = a.domain
group by d.id, d.domain
order by count(id) desc;
count | id | domain
-------+-----+------------------------
2 | 179 | computerworld.com
2 | 185 | sco.com
1 | 177 | aec.at
1 | 178 | coachella.com
1 | 180 | marca.es
1 | 181 | metica.se
1 | 182 | msn.com
1 | 183 | nickjr.com
1 | 184 | pason.com
1 | 186 | watchit.com
1 | 187 | www.pcbanking.cibc.com
or
agents=# select distinct a.agent
from agents a join domains d on a.domain = d.id
where d.domain = 'computerworld.com';
agent
----------------------------------------------------
Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1)
awesome. not sure what to do with the data yet, but i have some ideas. like ... generate kio_httprc add on files for people to download containing the top N sites; country specific kio_httprc add ons; a way to quickly check konqueror bugs for specific domains for known user agent quirks .... sounds like a nice application for HotNewStuff, too.
but to make it all work really well ... i need you! well... not really you, but your kio_httprc file where all these mappings are kept. the more of these files that get fed into the database, the better. so if you'd like to help out email me your `kde-config --localprefix`/share/config/kio_httprc file and i'll add it to the database.
(for the privacy conscious, it's all anonymized in the database. and if it makes you feel any better, i'll trade you your kio_httprc for mine upon request ;)

3 comments:
Heh. It's interesting how these things work out ;)
It was just a week or two ago that Opera 8 beta 2 was released, wherein Opera finally buckled down and decided to include an automatically-updated ua.ini file with a list of misbehaving sites and what user agent to use for them (they had been previously reluctant to do this) -- essentially the same thing you're now proposing/doing. (They also added 'report a site problem' to the help menu, which isn't a bad idea, though Konqueror already has 'report a bug'.)
Perhaps it might be worth a look at one of those ua.ini files (though mine seems to be empty, oddly enough), as I wouldn't be surprised if Opera and Konqueror both choke on (or more appropriately, get choked by) many of the same sites, both of them having rather low market share...
I can relate to that post. Finding the right domain name these days is more challenging and definitely more critical to getting good search engine positioning.
I really enjoy your blog, keep us posted.
This blog is awesome! If you get a chance you may want to visit this kazaa free download site, it's pretty awesome too!
Post a Comment