Daniel Stenberg wrote:

>> Thanks for the pointers. What would you like me to do next:
>> (b) generate a new patchset relative to origin/master; or
>
> That'd be my preferred action!

OK, see the attached patches. It took a long time to make small
changes, but my knowledge of git has expanded greatly :-)

I'm not sure if my removal of the special-case treatment of telnet
uploads is the optimal solution, but the combinations of:
* named-file versus stdin
* curl tool versus other libcurl user
* telnet versus other protocol
* blocking versus non-blocking read
* fread versus CURLOPT_READFUNCTION
were in danger of making my head explode! Anyway, my test cases pass
now where they failed before.

-- 
Colin