fetch extension - use local filename from content-dispositionheader (new diff)

Dag-Erling Smørgrav des at des.no
Sat Dec 31 06:02:16 PST 2005

Martin Cracauer <cracauer at cons.org> writes:
> Well, it's not you who started it.  DES called my code "disgusting"
> which by all measures to apply it was not.

Eye of the beholder, etc.

> The only reason why we have an ABI problem is that he chose that
> interface, an interface which does the best (worst) possible job of
> locking the ABI against future extensions to access more of the
> growing number of useful http headers in common use.

You forget that libfetch is not just about HTTP.

> Leaving aside for a moment that both clients of the library are in the
> FreeBSD base system and hence not affected by an ABI change, when I
> offer him to change that to an interface that is robust against future
> extensions (and can even be implemented API and ABI compatible to the
> current one) he cuts my quote to a half-sentense and adds a one-word
> reply.

Because you keep further and further in the wrong direction.

> So he limits his maintainership to "would like notification" and he
> has obviously seen the notification.

What would you like me to put in MAINTAINERS instead of "would like
notification"?  No matter what I put there, people will ignore it
(like you did), and stronger wording will only incite stronger flame

The fact of the matter is that libfetch and fetch already have way too
many interacting features.  Even small changes to one of them can have
unintended consequences for others (witness the repeated breakage of
-r and -m over the past two years).  Therefore, any non-trivial change
which does not further the primary purpose of these tools (namely to
support the ports collection and the package system) will be regarded
as suspect.

(BTW, your patch breaks -r and -m.  Ironic, isn't it?)

If libfetch had a mission statement, it would be: allow applications
to efficiently access ftp://, http:// and file:// URLs as regular C

If fetch had a mission statement, it would be: provide a command-line
interface to libfetch which the ports system can use to fetch

Your patch does not further any of these goals.  It does however
increase the complexity of both libfetch and fetch, threatening their
ability to fulfill their mission.

FWIW, what you want to do can easily be done in five or ten lines of
Perl using Net::HTTP.

Dag-Erling Smørgrav - des at des.no

More information about the freebsd-current mailing list