Re: [Gnu-arch-users] Features command for arch

[tomas]

On Thu, Sep 02, 2004 at 10:56:43AM -0700, Tom Lord wrote:
> 
>     > From: address@hidden
> 

[...]

> I like the word "introspection" here.   That *is* (i agree) the key
> thing.

[...]

> We have some rough ideas so far such as:
> 
>   ~ sub-command CLIs change over time [...]
>   ~ there's lots of "weird branches" of `tla' running, [...]
[...]
> That second point underscores for me the need for standards (I'm
> working on it, actively but slowly).   But about "features".....

OK.

> I'm still a little vague on what they are.  I guess the related
> question is "What does it *mean* for a `feature' to be present (or
> absent)'?" from which maybe we can figure out how to name features and
> how to implement `tla features'.

>     > [``provide'']
> 
> That is close to a design trap you are falling into there.
> 
> You can see some nice (rough) analogy to another system, so you posit
> the (underspecified) existence of some feature based on that analogy,
> then show how it would be fun to have, then convince everyone to go
> off and implement it.

Darn! You caught me! ;-)

>                       (I'm not accusing *you* of going overboard
> .... I'm just pointing out a trap that is easy for anyone to fall
> into.)

No, but maybe causing others to go overboard (half-kidding)

> The problem: nowhere in that process, between you thinking of the
> analogy and you convincing everyone to work on something, did you
> actually "validate" the analogy.   Possibly the analogy makes no
> logical sense and everybody's work will be wasted.
> 
> What does "add (provide 'signed-archives) to tla" actually,
> rigourously, actually *_!_mean_!_*?

Yep. That's the point we'll have to find out.

[Rest elided -- I agree]

Regards
-- tomás