Re: Three categories of log files

From: Balachander Krishnamurthy (bala@research.att.com)
Date: Mon, Apr 26 1999


Message-Id: <199904261022.GAA74124@raptor.research.att.com>
To: Henrik Frystyk Nielsen <frystyk@w3.org>
Cc: www-wca@w3.org
Date: Mon, 26 Apr 1999 06:22:30 -0400
From: Balachander Krishnamurthy <bala@research.att.com>
Subject: Re: Three categories of log files 


henrik asks for comments on 
> The discussion of certified vs. uncertified logs made me realize that there
> in fact are three categories and not just two:
> 1) certified entries which have passed a set of tests and we can recommend
> 2) entries that failed the certification test
> 3) entries that haven't been processed yet
let us take the above classification

> 1) should clearly be in the repository. 
> 3) obviously can't be in the repository 
agreed.

>but I strongly believe that 2) should be in the repository as well indicating:
	> 	- why they didn't pass
	> 	- how serious the errors are
	> 	- how they can be avoided in the future

disagree. the people who obtain logs from the repository have to get the
above list (why they didn't pass, seriousness of errors, avoidance) and
not ignore it. it is quite likely that people will take any logs in the
repository, passed or otherwise, and use them. they will cite the repository
as the place where they got it, and not necessarily the list of errors in
it. 

> This allows for the following:
> 	- other people can verify our certification
> 	- other people can learn from previous mistakes

good points. so just have the excerpts from the uncertified logs pointing
out the mistakes and not have the logs itself. so if log foobar has 11 fields
instead of 8, the timestamps are in the future, includes response code 445,
methods GETALL, then list those records and tell the supplier to fix them
if they want that log to be included in the repository and record all
this in a BUGS file. 

this way, others can verify our certification and learn from previous mistakes.

cheers,
bala