Greg Hudson <ghudson@MIT.EDU> writes:
> After making this change, I'm not a big fan of the way the dumpfile
> parser exports the parser interface. I doubt anyone will ever use
> that functionality, so it just makes it harder to add new features to
> the dump format.
[Responding to this bit separately from the patch itself.]
The point of defining the parser interface was so that svn dumpfile
format could be easily used as an export format to other systems.
Offering a programmatic export format is even better than offering a
static one. CVS essentially offers neither, and anyone working on
cvs2svn can testify to the pain :-).
One could respond: "It's easy to whip up a parser for our simple
dumpfile format, so why do we need to publicize the interface?" But
it's the sort of thing where people are likely to make little
mistakes. We shouldn't ask them to reinvent the wheel, when we can
just offer the wheel we had to implement for ourselves anyway.
Well, you probably knew all this -- I'm really responding so others
see the answer to your question. I think someday we (or someone) will
be very glad we exported this interface, though the only evidence I
can offer for this conjecture is our experience with cvs2svn.
> (Also, what's the point of exporting
> svn_repos_get_fs_build_parser, when the only thing you can do with it
> is call svn_repos_parse_dumpstream on it, and doing that is exactly
> the same as calling svn_repos_load_fs?)
That I can't answer :-(.
-K
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Mar 29 19:17:24 2004