Helder Geovane kindly asked me to post this here, after I published a comment at http://meta.wikimedia.org/wiki/Talk:SVG_image_support#Batik_server_revisited
This started out when a colleague posted a proposal on strategy wiki of throwing some foundation money at the active development of rsvg. I wanted to counter this with an old question: why not use Batik instead? The decision for rsvg in 2006 was based on the asumption that a Java application was a bad idea due to overhead. At this time, a user proposed a http Server app based on Simple (http://www.simplframework.org). He published a demonstration version (http://batikd.sourceforge.net/), but noone reacted, and his work stopped. Now I have taken a look at his approach, and I find the results are not that bad, for an early development stage: in most cases, the batikd rendering times are two or three times those of rsvg, and in some specialized cases (extensive use of filters), batikd is even faster.
I would say, like rsvg improvements, a Batik server for a production environment would need considerable work, and I am frankly not qualified to do it. So, regarding the funding proposal my position is to say that there are multiple avenues to get better SVG support. (I have seen the comments about svgweb, another project that seems to need work before it is usable.)
This post here is only meant as an encouragement if someone who knows more about server development than I do would like to take a deeper look.
On Wed, Aug 12, 2009 at 7:19 PM, hk_knghk.kng@web.de wrote:
Helder Geovane kindly asked me to post this here, after I published a comment at http://meta.wikimedia.org/wiki/Talk:SVG_image_support#Batik_server_revisited
This started out when a colleague posted a proposal on strategy wiki of throwing some foundation money at the active development of rsvg. I wanted to counter this with an old question: why not use Batik instead? The decision for rsvg in 2006 was based on the asumption that a Java application was a bad idea due to overhead. At this time, a user proposed a http Server app based on Simple (http://www.simplframework.org). He published a demonstration version (http://batikd.sourceforge.net/), but noone reacted, and his work stopped. Now I have taken a look at his approach, and I find the results are not that bad, for an early development stage: in most cases, the batikd rendering times are two or three times those of rsvg, and in some specialized cases (extensive use of filters), batikd is even faster.
I published some batch-batik performance numbers back when this was discussed. It wasn't too appealing.
I still hold the position that if we were to use anything other than rsvg it should be inkscape, since at least then users could have a hope of some degree of bug-compatibility WYSIWYG. ;) (and inkscapes performance is attractive... and at least all the crash bugs I reported were fixed :) )
2009/8/13 Gregory Maxwell gmaxwell@gmail.com:
On Wed, Aug 12, 2009 at 7:19 PM, hk_knghk.kng@web.de wrote:
Helder Geovane kindly asked me to post this here, after I published a comment at http://meta.wikimedia.org/wiki/Talk:SVG_image_support#Batik_server_revisited This started out when a colleague posted a proposal on strategy wiki of throwing some foundation money at the active development of rsvg. I wanted to counter this with an old question: why not use Batik instead? The decision for rsvg in 2006 was based on the asumption that a Java application was a bad idea due to overhead. At this time, a user proposed a http Server app based on Simple (http://www.simplframework.org). He published a demonstration version (http://batikd.sourceforge.net/), but noone reacted, and his work stopped. Now I have taken a look at his approach, and I find the results are not that bad, for an early development stage: in most cases, the batikd rendering times are two or three times those of rsvg, and in some specialized cases (extensive use of filters), batikd is even faster.
I published some batch-batik performance numbers back when this was discussed. It wasn't too appealing.
I did some a year and a half ago, they should really be redone:
http://www.mediawiki.org/wiki/SVG_benchmarks
I still hold the position that if we were to use anything other than rsvg it should be inkscape, since at least then users could have a hope of some degree of bug-compatibility WYSIWYG. ;) (and inkscapes performance is attractive... and at least all the crash bugs I reported were fixed :) )
Yuh. My tests it came out very slow, and not a good idea unless you're running Windows (since it comes in a standalone package) - but others have suggested that could be starting it afresh every time - if it could run from the command line with a single instance, it could be a bit faster, I don't know.
- d.
David Gerard <dgerard <at> gmail.com> writes:
I did some a year and a half ago, they should really be redone:
I knew I had seen the tests somewhere. I'll be running the files with batikd, and post the results. Due to time restraints, this will have to wait until the weekend.
wikitech-l@lists.wikimedia.org