[Grub-dev] do we even really need a native client

[Balinny]

jer wrote:
>> It's the <nl> at the beginning of doc the one not being honoured.
>>     
>
> Could you paste the section of the spec and explain?  I must be  
> missing it...
>   
Maybe i'm missing something.
If you add them with 
  print $arc "http://$host$path $ip 19691231175959 $ctype 
",length($body),"\n$body";
The file will look like:

http://www.example.com 127.0.0.1 19691231175959 message/http 11\nHello Worldhttp://www.example.net 127.0.0.2 19691231175959 message/http 7\nGoodbye

Which would correspond with <doc><doc>
And each of <doc> would consist of <URL-record><nl><network_doc> 

Whereas it should be <nl><URL-record><nl><network_doc>

\nhttp://www.example.com 127.0.0.1 19691231175959 message/http 11\nHello World\nhttp://www.example.net 127.0.0.2 19691231175959 message/http 7\nGoodbye

You see \n before http: on the archives because almost every webpage ends with some line feeds. But they belong to <network_doc>  (are into the count).


>> Right. But a client could wish to speak HTTP/1.1 There's still that
>> problem  with the compression.
>>     
>
> The workunit specifies the full request, including the HTTP/1.0  
> part.  If a client is smart enough (and wants) to munge the requests  
> in the workunit into 1.1, it'll have to be smart enough to force the  
> responses back into 1.0, seems like some unnecessary work though?
>   
Probably for the benefit of having persistent connections, as the hosts 
to connect are spread.
However, getting the result compressed seems worthwhile.

>> What about the user-agent vs agent issue?
>>     
>
> Explain?
>
> Jer
The workunits have a header: "Agent: Grub WU1", but such header is -to 
my knowledge- not
defined anywhere. I think it was meant to be User-Agent (rfc1945 section 
10.15 / rfc2616 sect 14.43).