Re: The 7 Deadly Sins of Versioning

Andre van der Hoek wrote:

> Both provide a speedup that can be quite considerable, and both might be
> needed for everything to really work (consider that 1 meg image where I
> change one pixel, compressing helps, but boy, I would rather send the
> diff!).

There may be a snag here: if the image is compressed (as it probably is), then a
diff on the bitstreams may not be all that small, since changing that pixel may
change much/all of the downstream data.

One possibility would be to permit type-specific diffs; if you make small changes
in an image, you can send a diff that says "patch the image by changing this
region", sort of like the interframe compression in MPEG.  Similarly, if you've
got a compiled Mac application under versioning, and you change a single string
resource, you could send a diff that says, "change 'STR ' resource 1717 to
'foo'".

I'm not sure I'm actually advocating this approach; the complexity might not be
worth the bandwidth savings.  I would certainly not advocate putting any
type-specific diffs into the required protocol; I would say make each diff format
a negotiable extension.

--
/====================================================================\
|John (Francis) Stracke    |My opinions are my own.|S/MIME supported |
|Software Retrophrenologist|=========================================|
|Netscape Comm. Corp.      | This is the .sig that says...           |
|francis@netscape.com      |        Ni!                              |
\====================================================================/
New area code for work number: 650

Received on Monday, 8 June 1998 12:08:14 UTC