Handling of Negative Timeouts different than python-memcached.

[jacobb]

pylibmc currently treats negative timeout as long-lasting (assuming Default behavior of "Never"/0), while python-memcached has them expire immediately.

This should be modified to match python-memcached or have the difference noted in the documentation (currently giving me a bear of a trouble getting tests to pass adding pylibmc support into django 1.3 :))

[lericson]

Hm, what is the point in keys that expire immediately? This, I feel, should rather raise an exception and be noted in the documentation. (also can't you just max(time, 0) it?)

[lericson]

Lack of input, so duly noted in documentation.

[jacobb]

(bump)

It seems due to a change in memcache or libmemcache (I'm not sure which?), pylibmc is now behaving with the odd behavior of having negative timeouts auto-delete (set and immediately forget) keys. Should the docs be updated again, or an error be raised?

[jacobb]

Possible related changeset: http://blog.elijaa.org/index.php?post/2011/08/17/Memcached-1.4.7-Released

[lericson]

Not really sure what you're rooting for here. Do you want an exception to be raised?

[jacobb]

I think removing the doc note is good enough -- as bizarre of a functionality as having negative timeouts delete a key is, it's at least consistent across all interactions with memcache -- through libmemcached or not.

[jacobb]

This is now covered more in depth in #202 , closing.