[Bug: 21.5-b29] gnus can't send in 21.5.29 but in 21.4.22

[Katsumi Yamaoka]

Lars Magne Ingebrigtsen wrote:
> Katsumi Yamaoka <yamaoka at jpl.org> writes:

>> Done.

> Great; thanks.

>> The new function definition encodes every part in the same way
>> practicing when actually sending them in advance and then looks
>> for things like a MIME boundary.  So it may slow message sending.

> But it should be done -- otherwise we can't be sure of successfully
> sending these messages, so I think your fix is totally the right thing.

But encoded data made for only computing a boundary can be used
for sending, too.  So, the mml functions for sending may be
rearranged so as to encode all parts of a message first, compute
a boundary, and combine them.  It seems not easy to achieve, though.

>> In addition, it assumes that signing and encrypting will never
>> generate a boundary pattern in data.

> True.  Hm... is that possible?

I'm not quite sure of it.  I was only worried a user who has no
caching mechanism will be bothered by being prompted for
a passphrase twice a part if performing signing or encrypting
for computing a boundary.