On Mon, 26 Apr 2010, Jonathan Robie wrote:
On 04/26/2010 03:22 PM, Justin Ross wrote:
Shuttling generated docs through subversion seems unnecessary. Any script
that can check for a change in a generated document can just as easily
check for a change in a source document. The only difference after that is
performing the generation, true?
I discussed this with #asfinfra earlier today. They won't run our build
system on their server. Their system is happy to copy a generated file over
when we check it in, but not to build the system itself.
Here's an excerpt from our conversation:
<jrobie> i.e., instead of checking in the artifacts, i check in a makefile or
ant file that builds the artifacts, it has to run on my system, you run it on
the client and then copy the artifacts over?
<joes4> no
<joes4> you're not running arbitrary programs on our frontline webserver
<joes4> we don't even allow cgi
<joes4> telling me you won't blame me for your mistakes doesn't help
<jrobie> so installing, say, xslt, docbook, fop, doxygen, epydoc ... that's
not going to happen
<joes4> the public will always blame the asf
You can try to convince Joe ;->
I think avoiding running programs on the frontline webserver is perfectly
reasonable. However, I don't think that quite argues against a script.
We just need to figure out (A) where to run it and (B) how to schlep the
results. B is the more important one, since it's not too hard to scare up
some cycles.
If we can't find satisfactory options for A and B, then I agree that using
subversion is more attractive. (Not really attractive, per se. More like
disgusting in a way that I can live with.)
Justin
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
