[OSM-dev] Tirex vs Renderd

[Jon Burgess]

On Sat, 2010-04-17 at 10:21 +0100, Kai Krueger wrote:
> Hello,
> 
> with the recent appearance of tirex there are now at least two, with the 
> python version of renderd 3, but probably more implementations of a 
> rendering backend. All attempt to roughly do the same, i.e. take 
> requests from a webserver or other program and render them. So the 
> question is what is the future outlook for each and the relation between 
> them? I.e. is tirex supposed to replace renderd and tile.osm.org will 
> eventually switch? Or was tirex just something geofabrik threw together 
> for a particular client and renderd should remain the "default"? Are 
> they intended to solve the same problem and "compete for the same 
> market" or does each have their own niche for which they are optimized for?

I don't think tile.osm.org will be moving away from the mod_tile renderd
in the near future but if tirex or something else demonstrates that it
can do a better job then I could be persuaded to switch. 

> This is less of a question though regarding using renderd vs tirex but 
> more of a question of where the future lies and thus for which to 
> develop new features. It would be a shame and rather inefficient to put 
> in effort to develop features for something that turns out to be a dead 
> end and my concern is to "fracture" the not directly large pool of 
> developers for renderd/tirex.
> 
> Yes, sometimes competition can be a good thing and thanks to it all 
> being opensource one can fork and merge at will, but I would be 
> interested to hear how others involved see the situation and help 
> potential contributors to make a informed decision to which project best 
> to contribute to.

I see a place for both. Geofabrik have developed something which fits
their needs better than mod_tile. On one level it is slightly wasteful
to have multiple implementations, but on another it means that we may
open up to a broader range of C/C++/Python/Perl developers. Any complex
problem like map rendering is going to have several 'best' solutions
depending on your priorities for determining merit. 

Even if a new feature was developed using one system does not mean that
it won't be ported to run with another. For example, if I was working on
enhancing the flexibility of the map rendering I might start
experimenting with the python version and then move to a C
implementation once I was happy with the approach. 

I would recommend you to pick whatever works best for you.

The open source world is full of competing projects: Mapnik/Osmarender,
JOSM/Potlatch, Linux/BSD, Gnome/KDE, Emacs/Vi, Bash/Csh, Python/Ruby,
Debian/Ubuntu/Fedora. I believe this diversity encourages innovation and
is probably a natural outcome of the bazaar style development model.

	Jon