> For example, if mirroring is a more restricted version of read-only, 
> then --mirror --refresh-index wouldn't make sense, nor would --mirror 
> --rebuild. There will be some examples of this just as soon as the bug 
> fix for 2693 gets pushed back. The idea is, if you know at start-up the 
> options passed don't make sense, error then rather than waiting a while, 
> and erroring when you don't have permissions to update the index (for 
> example).

Okay.  I'll take a pass over this to see if I can put in some option
checking.

> I'd argue that eventually, we might not want an
> InvalidContentException to be a show stopper and that instead we'd
> remove the corrupted file and try again from a different mirror, but
> perhaps not.

We could do that; however, we'd want to permanantly remove the mirror
from the list of mirrors under the assumption that the corruption was
malicious.  I'm not sure if we want to do that yet, though.

> Fair enough, then how about something during start up with says in the 
> server log (once) "Starting as mirror, only x, y and z are supported 
> operations" or "a,b,c,d,e are not supported".

I'm not sure I understand the rationale for this request.  Are you
concerned that admins/users won't be able to tell what methods the depot
actually supports?  If so, I've fixed the /versions/ method so it will
correctly return the methods and versions that any depot supports.  If a
method isn't supported in a particular mode, and the depot is running in
that mode, then the method won't be in the list of available methods.
Does this address your concern?

Thanks,

-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to