random-state.net

Nikodemus Siivola

<< next | top | previous >>

December 10th 2005 #
random, December 10th 2005

Verdict: Silly Me

My recent ASDF API change proposal turned out to be not just a bad idea, but needless to boot:

  • Many people depend on the current order of ASDF:*CENTRAL-REGISTRY* traversal.
  • The similarity to PATH environment variable was pointed out, and how tail to head traversal would be counterintuitive.
  • Anyone needing the behaviour I was after can obtain it thru an existing mechanism: ASDF:*SYSTEM-DEFINITION-SEARCH-FUNCTIONS*.

As penance I promise to write new documentation on this aspect of ASDF.

More then one people pointed out how nasty ASDF:*CENTRAL-REGISTRY* is as a direct interface, and proposed alternatives.

While I'm tempted by thoughts of a more abstract interface I'm also afraid that the cat is is out of the bag, milk spilled, etc: ASDF is widely deployed, and as the flurry of activity my proposal created proved *C-R* is widely depended on to work just as it does now. I believe any abstract layer built on top of it would have a hard time remaining robust as long as people have and depend on direct access.

IMO this is something to bear in mind should ASDF2 ever appear on the scene.

Meanwhile, in SBCL-land

Those of you following the SBCL side of this will be interested to know that to keep things consistent the order of traversal of SB-EXT:*MODULE-PROVIDER-FUNCTIONS* will also remain unchanged.

What will change change is that in the future the SBCL-provided ASDF will add its own module provider to the end of *M-P-F*, so as not to trump the contrib-module provider.

For casual users this means effectively nothing will change, except that you can no longer accidentally load eg. the portable version of asdf-install.

People hacking on their local copies of SBCL contrib modules will have to change their setups slightly. Directions for this will be added to the SBCL documentation.

No, Not The R-Word

I refuse to join the masses and say the R-word, but let it be said that we are definitely good in picking up the blame as a language-community. Some choise out-of-context quotes:

  • ...here's where the Lisp psychosis comes in...
  • ...if you can't deal with a little attitude.
  • ...unnecessarily emotional defenses of Lisp...
  • ...critique of the Lisp community which has a tendency to navel gaze...

Sigh. Can we go back to our regularly scheduled techical discussion, please?

Favourite HTML-element of the week

<blockquote>