Noel J. Bergman wrote:
What is the rationale behind the decision to keep the generated-site
under our version control?
Management and particularly disaster recovery by the infrastructure team,
which cannot and will not rebuild the site for every project using various
tools, nor wants to expend the resources (time and computing) while trying
to recover.
Ok. It makes sense, even if it would be better to have a simple backup
than a versioned repository for this.
I agree that keeping the versioning for generated artifacts is a waste
of space in Subversion.
The top culprits in our now 111MB site (up from 18MB for JAMES v2.1, of
which 11MB had been javadocs) are:
111M /www/james.apache.org
35M /www/james.apache.org/javadocs
29M /www/james.apache.org/javadocs/org/apache/james
2.5M /www/james.apache.org/javadocs/org/apache/mailet
61M /www/james.apache.org/server
4.4M /www/james.apache.org/server/xref-test
4.2M /www/james.apache.org/server/xref-test/org/apache/james
35M /www/james.apache.org/server/apidocs
29M /www/james.apache.org/server/apidocs/org/apache/james
2.4M /www/james.apache.org/server/apidocs/org/apache/mailet
18M /www/james.apache.org/server/xref
17M /www/james.apache.org/server/xref/org/apache/james
7.7M /www/james.apache.org/jspf
4.2M /www/james.apache.org/jspf/apidocs
3.6M /www/james.apache.org/jspf/apidocs/org/apache/james
1.8M /www/james.apache.org/jspf/xref
1.7M /www/james.apache.org/jspf/xref/org/apache/james
2.1M /www/james.apache.org/mailet
1.6M /www/james.apache.org/mailet/org/apache/mailet
5.0M /www/james.apache.org/rfclist
The javadocs for JAMES by itself are 3x than they had been in v2.1, and the
total size of the site is now roughly 6x what it had been. Now we may want
all of this, but let's not discount the explosion in space that just
occurred. Certainly not "just some extra (hundreds?) of HTML pages." The
entire site used to be roughly:
18MB = 3MB + 5MB (RFC docs) + 11MB (javadocs).
Now it is roughly
111MB = 10MB + 5MB (RFC docs) + 76MB (javadocs) + 20MB (cross-references)
You found that 111MB was not right. It was 42MB and now that I removed
the trash from the previous version we are to 29MB:
29MB = 1.7MB (website pages) + 2.3MB (RFC docs) + 15.5MB (javadocs) +
9.5MB (cross-references)
I just saw that the rfcdocs are in the wrong place. I have to fix it.
Yes, clearly the site itself is relatively tiny compared to the generated
artifacts.
But other than cross-references, there doesn't appear to be anything here
that would indicate that Maven is the cause of major bloat.
By the way, the javadocs linked as "James 2.3B javadocs" are the javadocs
for trunk (v3), and the ones at http://james.apache.org/javadocs/ (which
does not appear to be linked from anywhere) are for v2.3b3. And we have
three copies of the mailet javadocs: javadocs/, mailet/ and server/apidocs/.
--- Noel
I cleaned the old javadocs and the trunk docs is a known issue: I'll
work to fix it later. Btw thank you for the feedback!
Stefano
