Home freenode/#haskell: Logs Calendar

Logs: freenode/#haskell

←Prev  Next→ 502,152 events total
2021-04-21 11:30:07 <merijn> tatsumaru: There have been repeated attempts to create a new report
2021-04-21 11:30:07 <tatsumaru> https://mail.haskell.org/pipermail/haskell-prime/2016-April/004050.html
2021-04-21 11:30:19 <merijn> That's a mailing list message from 2016 :p
2021-04-21 11:30:40 <tatsumaru> so the effort flopped?
2021-04-21 11:30:42 <merijn> That attempt, like 4 or so others, has fizzled out
2021-04-21 11:30:56 <tatsumaru> is it due to lack of interest in the language?
2021-04-21 11:30:59 <merijn> No
2021-04-21 11:31:28 <merijn> The problem is that there is a mismatch between "stuff people want in the new report" and "stuff that needs to be done"
2021-04-21 11:31:51 <merijn> tatsumaru: Like, people want to include a bunch of GHC extensions into the report and make them "not extensions"
2021-04-21 11:31:58 kiweun joins (~kiweun@2607:fea8:2a62:9600:8080:6b3d:32d3:b207)
2021-04-21 11:32:03 <tatsumaru> i hope they don't turn it into the next pyhton-like monstrocity that's okay for everything and the best for nothing
2021-04-21 11:32:10 <merijn> The problem is that, *properly* standardising stuff is hard and boring work
2021-04-21 11:32:16 <tatsumaru> python*
2021-04-21 11:32:25 <merijn> tatsumaru: Like, just saying "whatever GHC is doing right now" isn't really a standard
2021-04-21 11:32:29 urodna joins (~urodna@unaffiliated/urodna)
2021-04-21 11:32:57 <merijn> tatsumaru: So someone has to sit down, figure out how GHC implements the extension, write out a specification of what it means, and figure out how it interacts with everything else
2021-04-21 11:33:14 <tatsumaru> accountant's work
2021-04-21 11:33:18 <merijn> tatsumaru: But most people just want stuff to "not be an extension" without doing the hard/boring work
2021-04-21 11:33:18 <__monty__> Part of the problem is also GHC's dominance. The reason to have a report is so multiple implementations can coexist. If there's not really any serious alternative implementations there's a lot less motivation for a report.
2021-04-21 11:33:42 <merijn> tatsumaru: Combined with the fact that GHC is basically the only *real* implementation, interest in standardising is low
2021-04-21 11:34:01 <merijn> If everyone keeps working on GHC, what's the point of an independent standard that no one else implements?
2021-04-21 11:34:12 ddellacosta joins (~ddellacos@ool-44c73afa.dyn.optonline.net)
2021-04-21 11:34:19 <tatsumaru> isnt GHC an implementation of haskell?
2021-04-21 11:34:23 <merijn> Yes
2021-04-21 11:34:48 <merijn> But GHC has much *much* more stuff than just Haskell2010 :p
2021-04-21 11:35:10 <kuribas> merijn: funny thing, Python doesn't have a report, but it has multiple implementations.
2021-04-21 11:35:19 <merijn> kuribas: Well, kinda, sorta
2021-04-21 11:35:20 <kuribas> which are also incompatible
2021-04-21 11:35:24 × rekahsoft quits (~rekahsoft@cpe0008a20f982f-cm64777d666260.cpe.net.cable.rogers.com) (Remote host closed the connection)
2021-04-21 11:35:40 <merijn> Python's spec is just "Whatever CPython does"
2021-04-21 11:35:43 <__monty__> That's what you get when you say "Implementation X *is* the reference."
2021-04-21 11:35:46 rekahsoft joins (~rekahsoft@cpe0008a20f982f-cm64777d666260.cpe.net.cable.rogers.com)
2021-04-21 11:35:50 <tatsumaru> scheme has a ton of implementations with new ones coming even recently, even though the userbase of the language is miniscule compared to haskell
2021-04-21 11:36:05 <tomsmeding> the language is far smaller
2021-04-21 11:36:15 <merijn> tatsumaru: That's because making a Scheme compiler is *much* simpler :p
2021-04-21 11:36:18 <__monty__> I think scheme's userbase is a lot bigger than you assume.
2021-04-21 11:36:29 × kiweun quits (~kiweun@2607:fea8:2a62:9600:8080:6b3d:32d3:b207) (Ping timeout: 250 seconds)
2021-04-21 11:37:18 <merijn> tatsumaru: GHC, for example, has a *ton* of optimisation stuff for Haskell code. If you have a cool new idea you *could* spend 5 years reimplementing that for your cool idea...or you could build it into a fork of GHC and save 5 years
2021-04-21 11:37:21 <yushyin> writing yet another scheme compiler is my weekend exercise
2021-04-21 11:37:37 <yushyin> :P
2021-04-21 11:38:41 <kuribas> yushyin: in haskell or GTFO :)
2021-04-21 11:38:49 <merijn> tatsumaru: So the compiler and language are alive and well. Efforts to standardise Haskell...not so much :p
2021-04-21 11:39:05 <tatsumaru> what's the benefit of standardizing haskell
2021-04-21 11:39:27 <kuribas> standardizing haskell == which extension to keep, and which to deprecate :)
2021-04-21 11:39:42 <merijn> tatsumaru: THat's unclear, which is why it isn't happening :p
2021-04-21 11:39:43 <yushyin> kuribas: various languages, it depends on my mood on that weekend ;D
2021-04-21 11:39:49 machinedgod joins (~machinedg@135-23-192-217.cpe.pppoe.ca)
2021-04-21 11:39:56 × ddellacosta quits (~ddellacos@ool-44c73afa.dyn.optonline.net) (Ping timeout: 245 seconds)
2021-04-21 11:39:59 <kuribas> yushyin: so "write a scheme compiler" is your hello world?
2021-04-21 11:40:06 <merijn> tatsumaru: Standardisation is good for keeping multiple implementations compatible, but there's only one implementation in mainstream use (GHC)
2021-04-21 11:40:25 <tatsumaru> should multiple implementations be compatible
2021-04-21 11:40:41 <merijn> Well, if they're not, they're not really the same language, no?
2021-04-21 11:40:51 <yushyin> kuribas: possible!
2021-04-21 11:40:52 <juri_> yeah, this is a painful situation. i'm feeling the pain, at least.
2021-04-21 11:41:02 <tomsmeding> it's at least nice to have a reference point, to be able to consistently say "here is where we comply, here is where we differ"
2021-04-21 11:41:03 <merijn> tatsumaru: Consider the C11 spec. It exists so that I know my C11 code is gonna do the same thing on each C11 compatible compiler
2021-04-21 11:41:16 <tatsumaru> well as far as I know python 3 and 2 aren't compatible, are they considered different languages
2021-04-21 11:41:24 <merijn> tatsumaru: I'd say do
2021-04-21 11:41:26 <merijn> *so*
2021-04-21 11:42:19 <tatsumaru> is this like the equivalent of FDA in programming?
2021-04-21 11:42:25 <tomsmeding> I guess you could say that C and Python are language families, whereas C89, C11, Python 2 and Python 3 are languages, if you want to get pedantic
2021-04-21 11:42:41 <tomsmeding> food and drug administration?
2021-04-21 11:42:44 <tatsumaru> yes
2021-04-21 11:43:05 <tatsumaru> by "this" I meant standardization
2021-04-21 11:43:26 <__monty__> Even the PSF considers them different languages. There's a reason there's a PEP that says python3 should never masquerade as a "python" binary.
2021-04-21 11:43:40 <tomsmeding> without a standard of what "Python 2" means, it's hard to talk about what exactly that language supports or not
2021-04-21 11:43:54 <tomsmeding> you can talk about what CPython 2 supports -- but that's something else than Python 2
2021-04-21 11:43:55 × knupfer quits (~Thunderbi@200116b82b5e7400e96d6e54df5761a8.dip.versatel-1u1.de) (Remote host closed the connection)
2021-04-21 11:44:02 knupfer joins (~Thunderbi@200116b82b5e7400a819040ab136d554.dip.versatel-1u1.de)
2021-04-21 11:44:41 <tatsumaru> so complicated...
2021-04-21 11:44:59 <tatsumaru> it seems futile to try to restrain all these decision trees to a singular specification
2021-04-21 11:45:39 <tomsmeding> a specification doesn't constrain, it standardises :p
2021-04-21 11:45:45 <tomsmeding> so that you can unambiguously talk about stuff
2021-04-21 11:46:12 <tomsmeding> a standard for C11 doesn't mean that your own home-brew compiler for a C-like language cannot exist
2021-04-21 11:46:23 <tomsmeding> it just means that you can say "here is how it differs from C11"
2021-04-21 11:46:49 <tomsmeding> without a standard of C11, you can only say "here is how it differs from what gcc 9.2.1 accepts"
2021-04-21 11:46:56 <tatsumaru> ah i see
2021-04-21 11:47:00 <tomsmeding> which is fine for some purposes
2021-04-21 11:47:35 Aquazi joins (uid312403@gateway/web/irccloud.com/x-fvdnxupudqhwtjny)
2021-04-21 11:47:40 <tomsmeding> but sometimes it's nice to have a central standard :)
2021-04-21 11:48:09 geowiesnot joins (~user@i15-les02-ix2-87-89-181-157.sfr.lns.abo.bbox.fr)
2021-04-21 11:48:23 <tatsumaru> is specification and standardization the same thing?
2021-04-21 11:49:58 <tomsmeding> when referring to programming language specifications, I guess yes
2021-04-21 11:50:14 tomsmeding wouldn't know the difference
2021-04-21 11:50:15 rodriga joins (~quassel@134.204.25.66)
2021-04-21 11:51:06 <tomsmeding> standard, oxford def. 2.2: "A form of language that is widely accepted as the usual form."
2021-04-21 11:51:21 <tomsmeding> specification, oxford def. 2: "A detailed description of the design and materials used to make something."
2021-04-21 11:51:24 <tomsmeding> ¯\_(ツ)_/¯
2021-04-21 11:51:43 <tomsmeding> I guess it includes less CS-specific jargon in the second case
2021-04-21 11:51:48 codedmart_ is now known as codedmart
2021-04-21 11:52:03 <tomsmeding> ah specification 2.1 is nicer: "A standard of workmanship or materials required to be met in a piece of work"
2021-04-21 11:52:12 <tomsmeding> what's a specification? a standard, among other things
2021-04-21 11:52:19 <merijn> See: "an ISO standard cup of tea" :p
2021-04-21 11:52:47 <tatsumaru> so one implies some level of quality control while the other one is just instruction?
2021-04-21 11:54:22 × nopf quits (~frosch@static.179.17.76.144.clients.your-server.de) (Quit: Lost terminal)
2021-04-21 11:54:24 <tomsmeding> dictionary definition != usage as technical jargon
2021-04-21 11:54:30 <tomsmeding> but maybe
2021-04-21 11:54:41 <tomsmeding> there's also "report" :p
2021-04-21 11:54:52 <tatsumaru> what tends to happen in the long term with languages when there's no standardization is it detrimental?
2021-04-21 11:55:37 <tomsmeding> depends on the people

All times are in UTC.