On Sun, Jan 9, 2011 at 2:17 PM, Mark Kettenis <[email protected]>
wrote:
>> Date: Sun, 9 Jan 2011 17:08:22 -0500 (EST)
>> From: Ted Unangst <[email protected]>
>
>>  #endif /* !SMALL */
>>               ftp_printf(fin, ssl, "GET /%s %s\r\nHost: ", epath,
>>  #ifndef SMALL
>> -                     restart_point ? "HTTP/1.1\r\nConnection: close" :
>> +                     "HTTP/1.1\r\nConnection: close"
>> +#else
>> +                     "HTTP/1.0"
>>  #endif /* !SMALL */
>> -                     "HTTP/1.0");
>> +                     );
>
> Doesn't this mean that servers that only support HTTP/1.0 will stop
> working now?

Only servers that don't actually follow the rules: a HTTP/1.0 servers
should accept HTTP/1.1 requests just fine, returning the HTTP/1.0
interpretation of them.  The version sent by each side is _that_
side's version, or rather, the version to which the request or reply
that it has sent complies.  c.f. RFC 2145.

If someone can find a server that is broken and barfs on getting a
request version greater than it expects, it would be interesting to
see how other common HTTP clients (curl?  netscape?  chrome?) behave
with it.  Do they all auto-retry with a HTTP/1.0?  Adding that, or a
manual option to force HTTP/1.0, to ftp would be annoying...


Philip Guenther

Reply via email to