You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Travis P <sv...@castle.fastmail.fm> on 2004/04/01 01:18:20 UTC

Deltification and Jars

I'm considering version-controlling finished Java jars that are  
promoted into production.  A jar is essentially a zip file.  I'm  
scratching my head wondering if I should build double-big jars with the  
no-compression option so that the Subversion repository will grow  
linearly with the size of the changes rather than rather having to save  
a big difference because of changes that cause a shift in compression  
making everything else appear different.  I'm likely to make numerous  
promotions with small changes.

I believe that compressed files would tend to require deltas that  
comprise the whole content of the file after the first change is  
encountered.  On the other hand, jars compress per-file and the files  
(class files and support files of text and graphics) tend to be quite  
small (currently my largest is < 30K).  So the changes are likely to  
remain small as most class files and support files are small.  This is  
different than if everything were, say, gzipped in a .tar.gz file.

I'm curious if anyone has already waded into these waters to consider  
the relative merits/demerits of using Subversion on zip/jar files and  
whether turning off compression is useful or not?  I imagine that this  
issue will come up for several people.

My current suspicion is that I can just use compression and not worry  
about it.  Using compression is my preference since the jars are  
distributed to users over networks via distributed filesystems.

-Travis

p.s. 1:
   Thus URL:  
http://subversion.tigris.org/subversion-dev/current/msg00158.html
   mentioned in a comment in
   in  
http://svn.collab.net/viewcvs/svn/trunk/subversion/libsvn_delta/ 
vdelta.c?rev=8016&view=markup
   is no longer valid.

p.s. 2:
    
http://subversion.tigris.org/files/documents/15/17/svn- 
design.html#Text%20Deltas
   mentions a paper.  Looks like the paper was updated and re-published  
as
   "Delta algorithms: an empirical analysis"
   ACM Transactions on Software Engineering and Methodology (TOSEM)
   Volume 7 ,  Issue 2  (April 1998)  Pages: 192 - 214


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org