libera/#commonlisp - IRC Chatlog
Search
15:42:04
prxq
Hi, I have an event queue, that passes events to lisp code (it's a telegram bot). When there's a bug, I'd like to have a restart for retrying the event / message. How does one do that?
16:08:29
jcowan
beach: You are not allowed to do anything with CLtL2 except the usual right to view it privately.
16:09:31
jcowan
The copyright page is visible on Amazon and makes it clear that all rights are reserved (note that just saying "All Rights Reserved" doesn't do this, and in fact is pointless boilerplate)
16:12:45
jcowan
"No part of this publication may be reproduced, stored in a retrieval systems, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher."
16:16:50
edwlan[m]
I wonder if there’s some way to work around that, given the recent ruling that the outputs of models like ChatGPT aren’t copyrightable
16:17:42
edwlan[m]
Some prompt such that the output isn’t considered a derivative, but it’s substantially the same content
16:45:46
gilberth
Hmm. I was pondering to offer CLtL2 along with my Nova Spec. I was just assuming it would be ok to do so because <https://www.cs.cmu.edu/Groups/AI/html/cltl/cltl2.html> exists.
16:47:51
gilberth
After all, I consider CLtL2 to be very well written and thought that some cross-reference from dpANS to CLtL2 would be useful.
16:48:35
pfdietz
Output of a model not being copyrightable doesn't mean the output isn't copyrighted. It could infringe on someone else's copyright. Filtering it through a LLM doesn't necessarily cleanse it.
17:24:11
jcowan
The output can't infringe on someone else's copyright unless (a) that copyrighted material was an input and (b) the output actually infringes (and the tests for "infringement" depend on the jurisdiction.
17:25:05
jcowan
gilberth: I don't think it's likely you'll get in trouble for providing the HTML-ized CLtL2 verbatim: there are already several well-known copies of it besides CMU's.
17:27:01
jcowan
pfdietz: The fact that two documents are identical doesn't mean either infringes the other necessarily. There has to be a causal connection between them.
17:29:36
jcowan
But it would infringe the LaTeX version because it is the result of a mechanical transformation of it. So the question would be, is it fair use.
17:52:57
jcowan
In US law, there are four factors that, when balanced against each other, guide courts in deciding what is fair use. Three favor you; "amount and substantiality" is against you, since you want to copy the whole book.
17:57:08
jackdaniel
if you are poor then you are filthy thief, otherwise you are a brave visionaire - that's my impression of the balancing thing
18:04:44
Bike
i was going to suggest writing the publisher for permission, but it looks like butterworth-heinemann is an imprint of elsevier now, so i'm even less optimistic than i started
19:02:36
gendl__
Hi, in atomics, it says on implementations (such as CCL) which do not have an atomic-push or atomic-pop operation, "a fallback based on CAS is available for usage" -- does anyone know how that works?
19:05:34
Bike
atomic-push = (let ((new (cons element nil))) (loop (let ((old READ-PLACE)) (setf (cdr new) old) (when (cas PLACE old new) (return new))))
19:08:05
Bike
also i just realized i can just link the actual source for gendl__. here's atomic pop https://github.com/Shinmera/atomics/blob/master/atomics.lisp#L112-L117
19:08:18
gendl__
pjb: how does with-lock-held prevent other code interleaving which is not bound by that *atomic-lock*
19:09:50
gendl__
Bike: Thanks. I must be working with an outdated version of atomics - mine does not seem to have atomic-push or atomic-pop on CCL. I thought that was just because of the platform I'm on.
19:10:24
Bike
they appear to have been added in the most recent commit - though that was two years ago. maybe your quicklisp dist needs an update?
19:11:55
Bike
hum, these macros don't proces subforms very well, but i guess without some implementation support for cas there's not much to be done on that front
19:17:43
pjb
gendl__: that said, you may wrap your data in objects and use methods to control access…