On Wed, 10 Sep 2003, David Byron wrote:

> We're doing some testing here of the --limit-rate option and something has
> come to mind. This may be obvious, but since I'm basically relying on the
> curl man page for the spec for this feature I'm curious whether you'd
> consider a patch to docs/curl.1 and probably docs/MANUAL that points out
> that the limit is on the amount of user data being transferred and does not
> take any protocol overhead into consideration.

I would certainly accept patches that make this clear(er), or rephrases of the
wording in ways that make the docs easier to understand.

> I am not suggesting any change to the implementation. I think it works fine
> as it is from looking at the code, but it is a bit hard to tell by looking
> at a protocol analyzer. At least until we dive into ethereal a bit more and
> learn how to ask it for the maximum number of http payload bytes transferred
> in one second.

I guess you should just trace a full transfer and use timestamps in the
logging and then run a script that checks the amount of traffic sent.

> As well, if anyone else has tested this feature, can they share the test
> setup?

When I implemented it, I just ran it with various numbers and tried
downloading very huge files from distance servers and I made sure that the
progress meter reported (roughly) the same speed I set the transfer limit to.

-- 
 Daniel Stenberg -- curl: been grokking URLs since 1998
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf