Greg Stein <gstein@lyra.org> writes:
> > Oh, I remember that. I'm just realizing now that it makes me all
> > quesy...
>
> What part? That an RA might need to store data, or that it gets stored in
> the WC?
Just that there are implementation-specific cookies the network layer
needs to store on the client side (the WC being the only place to
store it there, obviously). The distinction I'm making is between
data that _any_ conceivable implementation must use, such as revision
numbers and paths, versus data that only certain implementations use.
I'm not totally, unbearably quesy at it, just mildly quesy. I'm sorry
I haven't made some sort of non-vague response about this yet. I
thought I was going to work through the holidays, but various personal
obligations made that unrealistic, so I haven't been full-time this
week. :-( Will be able to think about this issue uninterruptedly on
Tuesday, though.
> I haven't even started on my crusade to get rid of the SVN equivalent for
> .cvspass. IMO, that information should be stored in the WC rather than
> depending on information associated the user who happens to be logged in and
> running the command. The SVN "rc" file can contain defaults, but the actual
> values used (could be different from the defaults!) should be stored in the
> WC. And since those are (also) RA-specific values, I'm not sure how we can
> avoid the basic concept of private RA properties. Now... where we put
> those... doesn't matter to me.
Oh, wow, *completely* with you on that one. Authentication
information should be stored in the WC that resulted from successful
authentication.
And this is a good argument for the unavoidability of storing
implementation-specific data (translation: "data Karl isn't intimately
familiar with") in the WC. :-)
Yes, I'm seeing your point...
Received on Sat Oct 21 14:36:18 2006