Subject: Re: libssh2 32500 bytes packet size limit

Re: libssh2 32500 bytes packet size limit

From: Daniel Stenberg <daniel_at_haxx.se>
Date: Fri, 11 Jun 2010 19:44:30 +0200 (CEST)

On Fri, 11 Jun 2010, Peter Stuge wrote:

>> The buffer size of the app doesn't really matter now
>
> I think they do, in the sense that if an app gives libssh2 a 64kb buffer
> then I think that should be the payload of a single SSH packet.

64K is too large for a single packet so either we'd make libssh2 create more
than one packet in a single call, or we have the current solution where it
takes as much as it can to make one packet out of it. I can't think of any
particular benefit with do the first way.

I think the biggest drawback with having libssh2_sftp_write() map to a single
FXP packet like this, is when the app uses a very small buffer as then libssh2
will send a small packet instead of merging many small into fewer larger ones
which may become a significant overhead. I think a decent approach to this is
to document it so that apps will use as large buffers as possible.

> Yep, and it'll work, but packet size is still stuck at the (32k-overhead)
> limit in the transport layer, which I think is the real problem.

But the 32-34K something is a real limit imposed by the specs, it's not
something we have invented. I'm afraid I don't understand what other way to
approach this that you are suggesting.

-- 
  / daniel.haxx.se
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2010-06-11