Chad Perrin wrote:
I personally felt we'd sufficiently discussed this to death, but now
there's 2 different folks who want to tear it apart some more. If
you're bored of this, tell me, and I will drag these folks either into
private discussions, or maybe onto the ports list. Tell me if you've
heard enough of this .....
Read below for my comments.
On Thu, Nov 15, 2007 at 10:56:12PM -0500, Chuck Robey wrote:
Chad Perrin wrote:
On Thu, Nov 15, 2007 at 03:34:26PM -0500, Chuck Robey wrote:
Chad Perrin wrote:
On Mon, Nov 12, 2007 at 08:23:23PM -0500, Chuck Robey wrote:
This makes a little file of descriptor words, but it's not set so a
regular editor can manipulate it; the special ports program is needed
to set or reset this list. All ports query this list in making the
decision as to whether or whether not to include a particular port as a
dependency.
Ugh. As far as I'm concerned, everything that pertains to system
configuration should always be human-readable and editable without
special tools. Trying to insulate things from human ability to directly
manipulate them tends to lead to rapidly increasing difficulty of
debugging configurations.
I might have agreed with this, except, I have lived for a good while
with the Gentoo "USE" lists, and I can tell you that having insufficent
control over what goes ontp those lists causes havoc both with the users
trying to select the proper wording of the lists, and the programmers
trying to decide how to have a particular USE keyword represent a
particular ports usage. You have to make certain that both users and
programmers have a definite, firm meaning in mind when they use the
keywords, because (in another's well chosen words) if you don't, USE
lists are a PITA. It takes firmer control of meaning to make certain
that the list doesn't devolve into that.
This is actual experience talking, in this case.
I don't see how that translates into "the user should not be allowed to
view what's going on behind the scenes in a text editor if (s)he wants
to."
I think you're becoming confused about who said what, because that
particular line (the last paragraph above) isn't anything that I wrote.
Quote:
This makes a little file of descriptor words, but it's not set so a
regular editor can manipulate it
That's the point I'm addressing. No more, and no less. The response I
received to addressing that did not seem to provide much support for that
quoted statement, so I let you know that I don't see how that translates
to "the user should not . . ." et cetera.
It's because, in actual experience with a system based upon usage of
keywords (a bit more compllicated than what I'm suggesting, but it IS a
real-life system, specifically Gentoo Linux. As someone else (I forget
who) said (and I fully agreed with him), "USE lists are a PITA. That's
true. I can't point with the same agsolute certainty to the reasons
it's a PITA, I think I know them, but the facts are as I stated.
Personally, I believe it's because the meanings of the keywords are
insufficiently standardized. That's my own opinion, but the fact that
maintaining USE lists is a PITA is fairly clear.
I want to move all the work of specifying the dependencies used by ports
from being done at build time to being done at system install time.
Further, I want to decouple the choosing of actual ports from
dependencies also ... I want users to say something like "I have no
audio", and this statement to be coded as NO_AUDIO, and all ports to be
guided by the settings of the list keeping this info. I have no name
for the lists, but I don't want to call them USE lists, because I'm not
suggesting we slavishly follow Gentoo on this, and using the same name
would give that impression. Maybe MACHINE_DEFS, something like that?
I'm not particularly good at making names.
A second part of this suggestion was a reject list of regular
expressions, and any ports matched would be ineligible to be built or
installed.
Lastly, my point about making sure that both the users and the ports
authors use the exact same meanings is, in my opinion, the detail
missing from the Gentoo implementation, so I'm proposing that the
maintenanace of the list be done thru a particular tool, which will
prominently display the actual meaning of the word being set. The only
reason to make the list binary is to force everyone to use the
(basically database technology) tool to manipulate the keywords, thus
stopping folks from misconstruing the meanings. That's my only reason
for that, and there are certainly other ways to go about it, so as long
as whatever is suggested requires folks to see the commonly accepted
definition when they set the list, I don't care how it's done. The list
could as easily be encrypted, I guess, that would also cause the same
work flow, in somewhat the same reasoning as we use for forcing folks to
use "vipw" to change the pasword list.
Please consider that we'll get another chance to argue this out when I
have the software ready, so we don't need to settle it now. I don't
want this to continue to pollute the -questions list.
At that point, I will prepare, in advance, use cases, all the
documentation, and the actual code, and everyone will get their chance
to rant and rave, alrighty? You can stop me cold, if enough folks don't
like the idea, that's how the development of FreeBSD goes, and I
wouldn't change a thing with that.
I'd rather that you produce software I want than software I don't,
though. That's why I tend to feel that it's better to sort out what is
and isn't wanted, why it is and isn't wanted, and both whether and how
that applies to what you propose to produce, before it's produced.
Obviously, I'm not saying that what I personally want should be the
driving force behind FreeBSD, but from where I'm sitting that's the
important part.
_______________________________________________
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
