[PATCH 0 of 1] RFC: SSH transport protocol upgrade

Sune Foldager cryo at cyanite.org
Sat Mar 13 10:01:34 CST 2010


To anticipate responses, I think the main bike-shedding issue is the 
name of the capability. I realize that we don't in general use version 
numbers when extending Mercurial, but this is kind of a special case: we 
are extending the low-level transport bits of the protocol - a thing we 
not need to do very often.

Also, using protocol=2 (or transport=2 or whatever), means we don't need 
to keep on adding new capabilities if we do change the transport layer 
again, but instead just replace with protocol=3, since we parse the 
protocol number at the client.

Another problem is figuring out a good short name for these features, 
and also the fact that the introduced features are different for HTTP, 
but I would like to call both capabilities protocol=2 to avoid having 
introducing discrepancy and since these changes actually unify the 
feature set of SSH and HTTP, making them more "the same protocol" when 
viewed from the layer above.

In short, I am not strongly attached to the name itself; maybe 
transport=2 is better; but I do believe we should number this change, 
lest we later need to introduce another capability with an even weirder 
name.

/Sune


More information about the Mercurial-devel mailing list