On Wed, May 21, 2008 at 11:05 PM, Bo Peng <[EMAIL PROTECTED]> wrote:
>> According to this,
>>   http://svn.haxx.se/users/archive-2008-03/0376.shtml
>> svn handles binary diffs efficiently. I suspect that it would likewise
>> handle zip -0 even more sensibly. Merging diffs may still be a problem.
>
> I think the issue here is not only about efficiency, but also about
> usability of diff. When you modify a few words in a base64  bundled
> .lyx file, 'svn diff' would be around these words, and you know what
> had been changed. When you svn a zip bundled .lyx file, 'svn diff' is
> likely to yield nothing informative.

Well diff -a somewhat informative, although very messy. I can clean
the diff of two zip -0 files as follows:

$ diff -a shfiles{,2}.zip | strings | grep '^[<>]'
<       top -n 1 | grep Cpu.s | grep -o '[^ ]*.id' #| sed s/[^0-9]//g
>       THIS IS BORKEN CHANGE top -n 1 | grep Cpu.s | grep -o '[^ ]*.id' #| sed 
> s/[^0-9]//g

ar is gives nicer diffs:

$ diff -a shfiles{,2}.zip | wc
     30     104    1154
$ diff -a {1,2}.ar | wc
      8      54     300

$ diff -a {1,2}.ar
164c164
< cooldown.sh/    1211341443  1000  1000  100644  469       `
---
> cooldown.sh/    1211382850  1000  1000  100644  491       `
175c175
<       top -n 1 | grep Cpu.s | grep -o '[^ ]*.id' #| sed s/[^0-9]//g
---
>       THIS IS BORKEN CHANGE top -n 1 | grep Cpu.s | grep -o '[^ ]*.id' #| sed 
> s/[^0-9]//g


It seems as if we could probably even merge the diff if different sub
files were changes. Still seems unsatisfying, as most likely in both
zip/ar files the base .lyx file would have been changed.

>> We could presumably save as zip if the user selects compressed, and save as
>> base64 if the user does not select compression. Then zip/unzip states would
>> be as the user expects from prior versions of lyx.
>
> This is not needed. In my proposal, plain text, base64 format is used
> without compression, and our existing compression feature is used to
> compress this file.

Storing embedded files as base64 text in the zip file would eliminate
most of the advantages of the zip format discussed.

-- 
John C. McCabe-Dansted
PhD Student
University of Western Australia

Reply via email to