Re: using files with .xls and .doc in CVS

[Rob Helmer]

On Thu, Feb 14, 2002 at 11:09:36PM -0500, Eric Siegerman wrote:
> On Thu, Feb 14, 2002 at 03:59:01PM -0800, Rob Helmer wrote:
> > I agree with the rest, but can your describe how the "-kb" setting
> > can be "lost"?
> 
> I floated a suggestion a while back that -kb should be a bit
> stickier; basically, it should "trump" sources of -k options that
> would otherwise be higher priority.  The idea seemed to meet with
> general (though not universal) approval, if I recall, but
> nobody's written the code.  Neither have I, of course, so this
> isn't a gripe, just an observation -- and perhaps a gentle nudge,
> should the idea happen to appeal to you :-)

Ok, I see your point. I usually don't use the -k option except
to do -kb for binaries.

> 
> > [...] store spreadsheets and other "office" documents in a
> > human-readable, diffable/patchable format.
> 
> Bear in mind that these two are not equivalent.  Things like RTF,
> and often Postscript, aren't particularly mergable, even though
> they're text.  The same might easily be true of XML-based
> spreadsheet files, if the application isn't a bit careful about
> when it decides to stick in a \n.

Same could be said for source code :P

I think markup is pretty readable diff'd, I do it with HTML
all the time. Unified or context diffs are especially useful.

> 
> Even straight text isn't easily mergeable, if it's word-wrapped.
> I'm sure we've all had to deal with that at some point :-(
> 
> Basically, unless text lines correspond, in some sense, to
> records, diffs (and thus RCS/CVS deltas) will be suboptimal, and
> merges will be more conflict-prone.

To tell you the truth, I find auto-merges to be rarely worth it,
even with source code. It's definitely true that while concurrent
development is great, communication is absolutely key. Especially
if you start adding lines above where somebody else is currently
working, the merge can sometimes be quite painful if you let
CVS do it.

Concurrent development aside, what makes diffs wonderful is the
historical reference to the text, and the ability to pull older
versions of text up without having to store every single version.
To me anyway :)



Thanks,
Rob Helmer