On Thursday 21 November 2002 01:19 pm, Daniel Vainsencher wrote:

> Ned, about the network issues you raised - I'm not sure we want to
> address all of these in 3.4. I think we should assume that "stable"
> in terms of a release means "significantly better than the previous
> final". If Squeak's http requesting code wasn't industrial strength
> last time, but the Mail related protocol classes were ugly too, I
> think we should be happy if we get nicer mail this version, and let
> in the rest as it becomes ready.
>
> I think Michael's mail stuff may be applicable to 3.4, but that
> depends on more testing (we don't want to find out too late the new
> SocketStream/NetNameResolver code breaks Scamper, for example), and
> Michael making a clean release of only what's tested.
>
> It's possible as far as we know that Michael wants to continue to
> work on the HTTP stuff, and integrate the whole thing later.
>
> I personally think it would be better to test and integrate what we
> have now - there'll be enough to do next version, too ;-)

My concern is this: one of the significant changes to 3.4 is the 
addition of SqueakMap. It's important that this be working 
flawlessly.

To that end, I think we should be able to deal with network errors and 
installation errors better. I'm tired of just getting "installation 
of XXX failed" when the real cause is "got a 404 when trying to 
download it" or "it tried to extend a class that didn't exist" or "it 
didn't extract from the zip properly".

One possibility is to replace the HTTP downloading code *just for 
SqueakMap*, but I think we can make it so that we can diagnose 
download errors better.

And I know that I can improve the SMInstaller error reporting.

-- 
Ned Konz
http://bike-nomad.com
GPG key ID: BEEA7EFE