httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Hartill <>
Subject Re: logging
Date Thu, 25 May 1995 11:11:27 GMT
> Because, if a logging process is reading stuff off a pipe, it has to be
> formatted *somehow*, and since some of the fields are unrestricted strings
> (in particular, the client's request line, which much be treated as such
> if we want to robustly handle buggy clients), those strings need
> to be delimited somehow.
> That's what we're discussing --- the format in which the "raw info"
> is *transmitted* to the logging process.

Hmmm, you lost me in those threads then.

I don't like this one-line-fixed-format transfer because it 
immediately restricts the amount of info being passed. If some 
browser suddenly starts sending a new bit of info, we can't 
accomodate it without changing the "standard". It'll also be
very difficult to get other server developers to conform to this, and
even more difficult to keep them in sync should new pieces of info
become available at a later date.

I think we should prototype the MIME format passing idea... pass
as much info as is available, along with headers which describe it.

It seems much better to dump annotated info to the logger (n.b. the 
annotation can be abbreviated), and make the standard be that 

 1. loggers read the data in MIME header fashion.
 2. loggers can't assume any of the headers will exist.
 3. a set of (server dependent ?) abbreviation mappings may exist.
        [we'd have to give "SERVER_SOFTWARE" a standard abbreviation.]
 4. loggers can do whatever they like after taking 1-3 into account.

Rob Hartill                  

View raw message