[Mailman-Developers] Dates again

[Ian Eiloart]

--On Tuesday, November 23, 2004 19:13:58 GMT -0800 Joe Rhett 
<jrhett at meer.net> wrote:

> On Sat, Nov 20, 2004 at 10:42:24AM +0100, Brad Knowles wrote:
>> 	If there were a way to effectively detect when a "Date:" header
>> was wrong and when it was okay, then I might be willing to allow the
>> system to correct the "Date:" header in those particular cases.  If
>> you've got a patch or additional code that can do that, I'd like to
>> see it, although I can't promise it would be accepted by the Mailman
>> developers for inclusion in an upcoming version.
>>
>> 	I can't speak for Barry, Tokio, or the other Mailman developers,
>> but I would be violently opposed to a change of this sort.
>
> Out of curiosity, am I overlooking something or are we over-generalizing
> the solution?
>
> My straw man (which may be wildly under-educated) is that we have no need
> at all to modify any headers, but instead to ignore the Date: header
> entirely for archive and digest purposes.  Leave the Date header there,
> but don't use it.  Instead use time of receipt.
>
> Now, go grab a 2x4 and clue me in...

Hmm, I think that this whole conversation is not seeing the wood for the 
trees. Why would anyone want to sort a list by date? The date of a posting 
isn't really relevant except:

1. So that we can understand the currency of the information. For example, 
I don't want to be reading 1998 postings if I'm looking for information on 
the latest version of - say - Apache. For that it doesn't matter much that 
a date might be a few hours out.

2. Location in a conversational thread. It is important to read threads in 
order. For that, the important thing is the time that the email was sent, 
not the time that it arrived - which may differ by minutes, hours or days. 
However, the better solution is to sort by threads using in-reply-to 
headers. Google's gmail, and Apple's Mail application do this nicely.

I think that development focus should be on displaying threads properly, 
not on "fixing" dates that the system can't possibly know are right or 
wrong.

-- 
Ian Eiloart
Servers Team
Sussex University ITS