[uf-discuss] [chat] Microformats are not for data storage

[Colin Barrett]

On Oct 30, 2006, at 5:03 AM, fantasai wrote:

> Stephen Paul Weber wrote:
>>> Well, as I did when we first spoke months ago, I firmly disagree,  
>>> and
>>> over time, I think you will be proven wrong. There's simply no point
>>> in having multiple instantiations of the same data in a text-based
>>> format (I'm exempting relational databases).
>>>
>> I agree.  Universal formats are very useful, and if uF is being used
>> by one aspect of the program, why not by all?  I currently use XOXO  
>> to
>> backup my reading list / transfer it to other feed readers.  Why
>> should this not be able to apply across the board to all  
>> microformats?
>> It may not be part of 80% or initial-draft use cases, but if it
>> starts happening, I'll be all for it.
>
> Microformats are great for combining data and presentation. They're
> an interchange format that is optimized for that.
>
> But the back-end store of chat logs isn't about presentation, it's
> about data. And microformats are not optimized for data alone.
>
> Different chat clients display logs differently. The same chat client
> can have many different options for displaying logs: for example,  
> whether
> to show timestamps, what time zone they should be in, whether they're
> attached to every message sent/received or just stamped every five  
> minutes
> as a separate line, how to format the date, how to format the time,  
> etc.
> The log storage format should be one that is optimized as a data  
> source
> that can be easily transformed into any presentation. It shouldn't be
> optimized for only one presentation, as it would be if it mixed data  
> and
> presentation.
>
> Logfiles are better as XML than as relational tables because they're
> ordered structured data, because one doesn't do sophisticated  
> lookups on
> them, because they should be easy to parse and easy to index with  
> other
> tools.
>
> They're better as a standardized XML format like ULF than as hChat  
> because
> ULF is more compact; is easier to write, parse, and transform; can  
> easily
> include invisible metadata that affects presentation (like sender  
> status);
> and doesn't include extraneous presentational information.
>
> If the concern is being able to open up logs and look at them, I  
> suggest
> having the logs link to a helpful CSS file for direct presentation and
> an XSLT sheet for transformation to hChat.

Extremely well said, fantasai! A very nice summary of the arguments  
for an XML based log format.

Only thing I might disagree with is the SQL point -- sophisticated  
lookups can be done, especially when full-text indexing is on. But  
that's a whole 'nother ball game (one that was debated on the Adium- 
devl list for a number of weeks).

-Colin