Anderson Farias wrote:

> Hi,
>
>> Introducing additional commandline switches just because you don't want
>> to do away with old behavior is IMO wrong. In human-computer interaction
>> design they say that usually more choice for the enduser is usually a
>> bad design decision (and indicates a lack of willingness to make 'tough'
>> decisions).
>>
>> Either change the behavior of this option or keep the option as it is,
>> introducing yet another commandline switch is in my opinion unnecessary.
>
> Yeah.. sure.. so everyone that uses the -s option today will have it's
> applications/scripts broken when update, not to mention that IF changing the
> actual behavior *will add* new functionality anyway, unless it would not be
> possible to get full stats (user+system tables) anymore.

That was not exactly my point. Adding new functionality is not bad in
itself, but adding new functionality while also keeping the old,
similar, functionality usually leads to design by accretion.
Changing the behavior of a switch might break some things and can
therefor be considered a 'tough' decision, but adding yet another switch
to preserve backwards compatibility leads to an ever expanding choice
and thus to expanding confusion.

BTW: From what I understood from the development list, the option gstat
-a -s could be used to get the same output as gstat -s currently gives.
--
Mark Rotteveel