Hi, Bruce Stephens wrote:
I doubt that would be workable.
Yeah, that was where I'm in agreement with you.
I was suggesting deliberately keeping the hash the same, but allowing the thing refered to by it to be something other than the contents. (It would need to be something special, otherwise you'd end up regarding it as corruption since the hashes wouldn't work.)
Good point. Being able to differentiate between random corruptions and intentionally skipped files is important. (Intentional, in the sense that monotone noticed it doesn't have permission to retrieve the file in question).
Sounds very suspiciously similar to the sentinel revisions we plan to introduce for partial pull, no?
So all repositories would have an object addressible by a specific hash, but only in some of those would it be the original file.
Yes, I think we are on the same line of thought, here. Regards Markus _______________________________________________ Monotone-devel mailing list Monotone-devel@nongnu.org http://lists.nongnu.org/mailman/listinfo/monotone-devel