[Tilesathome] T at H vs. BOINC - Neverending story?

[Frederik Ramm]

Hi,

Philip Gillißen wrote:
> You're definitely drawing a black picture of TaH. It's really
> disappointing to read all these points

I often defend TaH when people talk bad about it. There was a time then 
TaH was cutting-edge, where it would deliver new renderings in as little 
as 15 minutes whereas the Mapnik server would take 15 days or more. TaH 
  really raised the bar back then, and without TaH we might still be 
stuck in a setting where you'd have to wait for days or weeks to see 
your mapping.

So TaH definitely has a place of honour in the history of OSM and nobody 
can take that away. But I don't think that it still is very useful today.

> All this consideration lead me to one question: Is TaH dead?
> It looks like the "at Home" addition is lost and does not create any
> advantage.
> A new infrastructure for "TaH v2" would be necessary. Is it too radical?

I think you're approaching this from the wrong side. The question 
shouldn't be: "We have all these cool kids with 20 cores each eager to 
contribute to something, what could they do for us?" - The question 
should be: What's a cool thing we could do for OSM which actually 
requires so much computing power that it cannot be easily done by a 
central rendering instance like tile.osm.org or bing or MapQuest?

It is possible that there are avenues of cartography not yet explored, 
for which something like the TaH infrastructure would really be useful. 
I'm thinking along the lines of

http://wiki.openstreetmap.org/wiki/TopOSM

- a beautiful map with a very complex rendering process (described on 
http://wiki.openstreetmap.org/wiki/TopOSM/Details), so complex indeed 
that it cannot currently be rendered "live". However, such maps require 
data in addition to what is on OSM, most importantly elevation data. 
Elevation data is bulky and probably unsuitable for downloading on the 
fly; it could be that certain clients would have elevation data for 
certain areas, and then pick up only rendering requests for their 
particular area.

But that's just my two cents. TaH does have some advantages over the 
centralised rendering, mostly that it has a very permissive styling 
which often brings ridicule from others but it is closer to the spirit 
of OSM than a centralistic approach. But technologically it really is a 
big waste of CPU cycles, and thus it should not be used as a prime 
example for distributed computing.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"