[jacquesm]

A good one, thanks. Yes, that makes good sense. I guess that in some languages twitter really does carry a lot more information than it does in Latin alphabet languages.

If your language would only allow for 140 bytes you'd be really out of luck if your language routinely requires multi-byte sequences in UTF-8.

Hm. You could do a 'tweet compression' trick where you use a fixed number of bit from a subset of UTF-8 that you know is multi-byte (selected for the extra long sequences) in order to put longer messages on twitter, and then use a decompressor to turn it back into ascii.

[madeofpalk]

Already exists! http://pages.physics.cornell.edu/~aalemi/twitter/