sdcv - commandline stardict client

marc zonzon marc.zonzon at gmail.com
Sun May 30 10:25:04 EDT 2010


On Sun, May 30, 2010 at 5:32 AM, Wolfgang Spraul <wolfgang at sharism.cc> wrote:
> Guylhem,
>
>> I'd suggest something close to
>> http://users.softlab.ece.ntua.gr/~ttsiod/buildWikipediaOffline.html -
>> it's simple and it works.
>> Trying to create zim files is complicated. (I tried!)
>
> Totally agree with you. Thanks for the link, seems like an interesting
> approach maybe we can get something like that to work quickly...
>
> For completeness, on IRC Kristian Paul pointed to a similar project:
> https://launchpad.net/wikipediadumpreader
>
> Wolfgang
>

On Sun, May 30, 2010 at 12:13 PM, jirka <jirka at penguin.cz> wrote:
> Wolfgang,
>
> It is a well-working software (client is written in PyQt). I used it on my
> Nokia N800 for
> some time. But it realy is not fast - even on the N800 with its huge RAM
> (128 MB) and
> relatively fast processor (400 MHz ARM).
>
> So it is probably too big and too slow for use on your little Bens.
>
> Probably it will be necessary to write custom software for this purpose to
> get satisfying
> performance (no, I don't want to try do it - I have lot of work with
> rewriting of my own
> software for use on the Ben :-\  ).
>
> Jiri Brozovsky

I have had a look on these software that I have tested on my laptop.
Both use a simple wikipedia dump that is free and easy to obtain that
they index; a quite cpu intensive task but that occur only once and
can be done on any server, not BNN. Then they use the index and a
frontend to query the database.
Wikipediadumpreader is written in python QT4 and, as say Jiri, it has
quite large footprint, on my desktop it take 38 Mega bytes of resident
memory. I don't think you can reuse it without rewritting the
frontend.
The earlier work of Thanassis Tsiodras "offline wikipedia reader", is,
along the author, a quick hack done in two days. The actual display
done by using raw mediawiki code in php. The search is done in xapian,
and the user interface is provided by a Django wsgi server. Of course
Django wsgi stand alone take 14M, but this layer is very thin. 60
lines of python mainly external call to commands for search and
display. Any piece could easily be replaced, but the Media wiki code
in php.
As it can be read by any browser (on BNN elinks, links2, lynx or w3m)
there is no problem of interface.
We could think to run Django as CGI, not standalone to avoid the big
daemon, but it would not enhance the memory and cpu load at time of a
request.
The power of Django to only call  two commands is clearly overkill, it
would be sufficient to rewrite this in whatever language, php, lua
being probably the lighter easy alternative.

Marc




More information about the discussion mailing list


interactive