I'm working on enhancing mailcrypt (support pgp 6.x, "better" remailer
support, blah blah) and I'm trying to understand why, at the start of
gnus-summary-edit-article the function (gnus-summary-show-article t)
is called.

I need to decrypt a message (perhaps) multiple times recursively.
Each successive decryption is applied to the current article.
mailcrypt calls gnus-summary-edit-article in order to do this.

To complicate matters, I use nnimap as my backend.  This backend does
not support writes.  So, rather than calling
gnus-summary-edit-article-done mailcrypt calls
gnus-summary-edit-article-postpone.  I suppose if the -done function
was called (and the backend supported editing) then the decryption
would be saved to the message store and subsequent calles to the base
-edit function would behave as expected.

Anyway, using -postpone, the next call to gnus-summary-edit-article
causes the article to be refreshed from the message store.  End of
recursive decoding since we'll allways be decoding the original
message rather than each intermediate step of the decoding process....

I'm not sure what would break if we didn't reload the article before
calling the -edit base.  Anyone know?

Attached is the "patch" if you can call it that.  I'd appreciate
knowing why this should not be applied...