On Apr 15, 2019, at 5:16 PM, Derek Atkins <derek(_at_)ihtfp(_dot_)com> wrote:
Jon,
I would rather 256K not be the One Required Size. It's larger than the
RAM size on some of my systems. I would still prefer 16K (or even 8K).
Bart, Neal? What do you guys think?
I spent part of the afternoon verifying GnuPG's chunk size. I started off using
--list-packets to get the size of something, and then moved to using pgpdump.
gpg --list-packets reports the combined length of all the partials as if it was
one big thing, but pgpdump shows different. I found that out after I sent my
last thing. I thought it was using single-chunk lengths even with some
moderately large files. Only when it was showing single size for something over
a gigabyte did I go grab pgpdump.
Playing around, GnuPG aggressively uses 8K chunks, and even when I encrypted
something below 8K in length, it used a 4K chunk, a 2K chunk, and then tied off
the rest.
Why not pick 8K?
Arguments in favor:
* It's what is done now.
* As noted in Bart and Neal's comments, there isn't going to be a huge
performance difference.
* It works in small implementations, as Derek says.
Arguments against --
* I have no idea. Someone who isn't me should say something even like "I want
it."
Anyone else?
Jon
_______________________________________________
openpgp mailing list
openpgp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/openpgp