Names are not descriptions; descriptions are not names

Names are not descriptions; descriptions are not names

[Adam Porter]

There is no better way to doom software to obscurity than to give it a
description as its name.  Names are not descriptions; descriptions are
not names.

Rarely, for a simple enough package, a short description may be a
reasonable choice for a name.  For example, "org-sticky-header"
conveys that it's for Org mode and that it provides a sticky header.
Assuming, that is, that the user knows what a "sticky header"
is--otherwise, what should it be called?
"org-header-line-that-shows-the-outline-path-at-the-top-of-the-window"?
Now imagine one user trying to share that package with another user:

     Alice: "Hey, I found a package that I think you'll find useful:
     org-header-line-that-shows-the-outline-path-at-the-top-of-the-window."

     Bob: "Oh, that does sound useful.  What's it called?"

     Alice: "I just told you."

     Bob: "Yes, but what's the name of the package?"

     Alice: "That's it."

     Bob: "No, I need the name so I can install it and try it."

     Alice: "I just told you the name."

     Bob: "No, you told me what it does."

     Alice: "Yes, but that's the name."

(With apologies to Abbott and Costello.)

I'm not merely joking, I'm also serious.  This is a real problem.

It's doubly a problem for a package that has similar functionality to
existing packages.  For example, with regard to "devil", the package
recently proposed by Susam Pal, various alternative, ostensibly
descriptive names have been proposed, like "key-transl",
"no-modifier-mode", "prefixless-mode", "implicit-ctrl-mode", and
"comma->control-mode".  If one of those names were used, how would a
user distinguish such a package from other packages that do similar
things, such as viper, evil, god-mode, meow, boon, and the numerous
other similar ones that are out there?  (Note that, although I use
none of those packages, I remember them because of their distinctive
names; had they descriptions as names, I would not remember them, nor
would I be able to easily find them again.)  We're faced with the
same problem:

     "Hey, I found a useful key-translate package.  You should try it."

     "Oh, I already use evil-mode."

     "No, not that one.  This one."

     "Well, which one is it?  There are a bunch of those, like viper,
     evil, god-mode, meow, boon..."

     "key-translate."

     "Yes, I understand that it does that, but what is it called?"

     "The name itself is key-translate."

And that brings us back to the heart of the issue: "Oh, ok.  So how is
it different than viper, evil, god-mode, meow, or boon?"  That is, the
user must read the description (or even the whole readme) to
understand what the package does.  The name is not enough; the name is
never enough.

And by burdening the name with a responsibility it cannot bear, the
name suffers, the package suffers, and ultimately, the user suffers.
The "descriptive" name is not memorable; the user likely forgets what
it's called a few weeks after installing and configuring it (who
hasn't experienced this already: installing a package, configuring it,
and then forgetting about it, finally being unable to remember the
name of the package that does the thing that one suddenly needs the
functionality of again, having to search ELPA or MELPA for such a
tool, and then discovering that one already has it--well, maybe it's
just me).

Of course, it's a laudable goal to reduce confusion for users.  Yet,
although Ed is the standard editor, do we not use Emacs?  What do
potential users say about that?  "Didn't Apple stop making those a
long time ago?"  Are they not confused?  Should we rename Emacs to
gnu-lisp-machine-with-built-in-editor?  Would we not shorten such a
name to GLiMBiE?  And would such a name be descriptive?

I'd like to share a link to a short essay published earlier this year
that reminded me of these threads on emacs-devel:
https://ntietz.com/blog/name-your-projects-cutesy-things/ Although
it's mainly about service names rather than software packages, its
points are relevant here:

     And then the cherry on top, the final nail in the coffin of
     descriptive names: They're just too hard to say and remember, and
     they're no fun. I don't want my services or projects to sound like
     a law firm ("Ingest, Processing, & Storage LLP"). A descriptive
     name will be wordy, or boring, or both. It won't be memorable, and
     it won't be fun. On the other hand, something that's cute will be
     far more memorable and much easier to say.

     The world is boring enough as is. Let's add more whimsy and
     cuteness through our service and project names.

As an Elisp package author, I can vouch that part of the joy of
programming is creation, and part of the joy of creation is in naming
one's creations.  Let us not steal this joy from the authors who
generously contribute their time and work to the public good that is
Emacs and ELPA.

I'll close with these words from Alan J. Perlis, quoted in Abelson's
and Sussmans' classic, /Structure and Interpretation of Computer
Programs/:

     I think that it's extraordinarily important that we in computer
     science keep fun in computing. When it started out, it was an
     awful lot of fun. Of course, the paying customers got shafted
     every now and then, and after a while we began to take their
     complaints seriously. We began to feel as if we really were
     responsible for the successful, error-free perfect use of these
     machines. I don't think we are. I think we're responsible for
     stretching them, setting them off in new directions, and keeping
     fun in the house. I hope the field of computer science never loses
     its sense of fun. Above all, I hope we don't become
     missionaries. Don't feel as if you're Bible salesmen. The world
     has too many of those already. What you know about computing other
     people will learn. Don't feel as if the key to successful
     computing is only in your hands. What's in your hands, I think and
     hope, is intelligence: the ability to see the machine as more than
     when you were first led up to it, that you can make it more.

Re: Names are not descriptions; descriptions are not names

[Jim Porter]

On 5/11/2023 9:01 PM, Adam Porter wrote:
> And by burdening the name with a responsibility it cannot bear, the
> name suffers, the package suffers, and ultimately, the user suffers.
> The "descriptive" name is not memorable; the user likely forgets what
> it's called a few weeks after installing and configuring it...

I strongly agree. While some projects have bad names (as an illustrative 
example, CHEATING: "obsCure and awkHward usE of lettArs Trying to spell 
somethING"[1]), a name's primary purpose is to be a memorable way of 
referring back to something you already know about. That means it should 
be easy to see how the name relates in some way to its purpose, but it 
doesn't need to be immediately apparent what it does just by looking at 
the name.

Probably my favorite project name in existence is "pacman", Arch's 
package manager. If I didn't know about it and you asked me to guess 
what it did, I'm sure I'd say, "It lets you play Pac-man," but once I 
know it's a package manager, the name makes perfect sense and I can 
easily remember it.

When I'm looking for a package on ELPA (or built into Emacs) to do XYZ, 
I don't generally consult the name at all, and instead read/search the 
description or the manual. About the only time I look at the name is if 
I'm searching for a major mode of a programming language, which is 
almost always "lang-mode".

That's not to say there are no issues whatsoever with naming: if a name 
is really obscure or particularly rude, those are good times to suggest 
a rename. For "devil" in particular, I think it's a pretty reasonable 
choice, since just by looking at the name I guessed that it was 
something in the same overall genre as evil. Maybe there's a better name 
out there, but I think it's still better than any suggested alternative.

However, I think there's potentially an even bigger issue: package 
discoverability as a whole. I imagine we all hope that Emacs will 
continue to get many new users, and they likely won't be able to guess 
anything about a name like "devil". We should do our best to make sure 
packages like that are easy to discover if you use some common search terms.

This might also include making it easier for novices to learn *how* to 
perform good searches. I'm not sure exactly what this would entail (it's 
hard to unlearn what you already know), but I am sure that it would be 
valuable. Emacs gives you an awful lot of tools to be able to learn on 
your own and dig deeper into customizing/extending it, but it can be 
hard to get started: it took me more than a decade to move past the 
stage of "copy-pasting Elisp code from other people's configs".

- Jim Porter (no relation)

[1] https://www.bmj.com/content/349/bmj.g7092

Re: Names are not descriptions; descriptions are not names

[Dr. Arne Babenhauserheide]

[-- Attachment #1: Type: text/plain, Size: 3073 bytes --]


Jim Porter <jporterbugs@gmail.com> writes:

> On 5/11/2023 9:01 PM, Adam Porter wrote:
>> And by burdening the name with a responsibility it cannot bear, the
>> name suffers, the package suffers, and ultimately, the user suffers.
>> The "descriptive" name is not memorable; the user likely forgets what
>> it's called a few weeks after installing and configuring it...
>
> Probably my favorite project name in existence is "pacman", Arch's
> package manager. If I didn't know about it and you asked me to guess
> what it did, I'm sure I'd say, "It lets you play Pac-man," but once I
> know it's a package manager, the name makes perfect sense and I can
> easily remember it.

That name is descriptive, but unique enough to latch onto memory.

And you can discover it if you remember "something about packages", so
you type pac<TAB><TAB> on the shell.

> When I'm looking for a package on ELPA (or built into Emacs) to do
> XYZ, I don't generally consult the name at all, and instead
> read/search the description or the manual. About the only time I look
> at the name is if I'm searching for a major mode of a programming
> language, which is almost always "lang-mode".

I fulltext-search the package listing.

    M-x package-list-packages C-s <whatever> C-s (repeat)

I don’t have time to read all the one-line descriptions.

And I like having somewhat descriptive names.

Try to remember how to open a PDF on the command-line. Gnome: evince —
it took me months to remember that. KDE: kpdf. It once was. Now it’s
okular. That’s still somewhat evocative of seeing something, so better
than evince, but I preferred kpdf. And for kwrite it is obvious what it
does. Same for gedit — I never forgot gedit, even though I use it
rarely. If I’m not yet fully awake in the morning when I get up to make
the lunchboxes for the kids, I can still remember gedit. I can also
remember icecat, because that’s like firefox. And lftp still works. But
audacity I have to search when I didn’t use it for a while. Same for
clementine and even worse "whatever is the default image viewer" (no, I
really don’t know; doesn’t help that the name is not in the program
listing ...) — and don’t ask me to remember xdg-open or M-x
browse-url-xdg-open — it always takes me several tries to remember the
correct command, and that’s with ido-completion in the commands.

Or worse: the advanced video-editor that’s not kdenlive.

Though naming with something else than the most obvious description
becomes important once there is more than one package for the same task.

I often use amx to search for something: M-x org-<something I want to do>

And all in all there’s an old truth: naming is hard. While a fixed rule
like "name not description not name" can help to avoid the trap of
thinking that the name *must* be a description, I think it is too
simplistic to steer the process of finding a good name.

Best wishes,
Arne
-- 
Unpolitisch sein
heißt politisch sein,
ohne es zu merken.
draketo.de

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 1125 bytes --]

Re: Names are not descriptions; descriptions are not names

[Po Lu]

Adam Porter <adam@alphapapa.net> writes:

> Subject: Names are not descriptions; descriptions are not names

They should be.

Re: Names are not descriptions; descriptions are not names

[Adam Porter]

On 5/12/23 01:04, Po Lu wrote:
> Adam Porter <adam@alphapapa.net> writes:
> 
>> Subject: Names are not descriptions; descriptions are not names
> 
> They should be.

Then shouldn't we rename Emacs?

RE: [External] : Re: Names are not descriptions; descriptions are not names

[Drew Adams]

> > Atom Porter writes:
> >
> >> Names are not descriptions; descriptions are not names
> > They should be.
> 
> Then shouldn't we rename Emacs?

Atom's already been named (and deprecated). ;-)


Oh, the treachery of names!
This is not a name.

Re: Names are not descriptions; descriptions are not names

[Po Lu]

Adam Porter <adam@alphapapa.net> writes:

> Then shouldn't we rename Emacs?

Its name is a historical accident, but if it were possible, then yes.
Unfortunately, `Editor' is already taken.

Re: Names are not descriptions; descriptions are not names

[Dmitry Gutov]

On 12/05/2023 16:58, Po Lu wrote:
> Adam Porter<adam@alphapapa.net>  writes:
> 
>> Then shouldn't we rename Emacs?
> Its name is a historical accident, but if it were possible, then yes.
> Unfortunately, `Editor' is already taken.

Editor#2.

Re: Names are not descriptions; descriptions are not names

[chad]

[-- Attachment #1: Type: text/plain, Size: 280 bytes --]

On Fri, May 12, 2023 at 2:06 AM Po Lu <luangruo@yahoo.com> wrote:

> Adam Porter <adam@alphapapa.net> writes:
>
> > Subject: Names are not descriptions; descriptions are not names
>
> They should be.
>

Bertrand Russel would _emphatically_ disagree with you.

~Chad

[-- Attachment #2: Type: text/html, Size: 701 bytes --]

Re: Names are not descriptions; descriptions are not names

[Eli Zaretskii]

> From: chad <yandros@gmail.com>
> Date: Fri, 12 May 2023 13:52:27 -0400
> Cc: Adam Porter <adam@alphapapa.net>, emacs-devel <emacs-devel@gnu.org>
> 
> On Fri, May 12, 2023 at 2:06 AM Po Lu <luangruo@yahoo.com> wrote:
> 
>  Adam Porter <adam@alphapapa.net> writes:
> 
>  > Subject: Names are not descriptions; descriptions are not names
> 
>  They should be.
> 
> Bertrand Russel would _emphatically_ disagree with you. 

Bertrand Russel cannot post here, as he is not subscribed.

Re: Names are not descriptions; descriptions are not names

[Philip Kaludercic]

Adam Porter <adam@alphapapa.net> writes:

> There is no better way to doom software to obscurity than to give it a
> description as its name.  Names are not descriptions; descriptions are
> not names.

I agree, but this is a strawman.  As you are referring to the thread
about "devil-mode" from earlier this week, I'd argue the position was
not to find a name that perfectly describes the package, but to find a
name that is remotely related to the functionality of the package.

> Rarely, for a simple enough package, a short description may be a
> reasonable choice for a name.  For example, "org-sticky-header"
> conveys that it's for Org mode and that it provides a sticky header.
> Assuming, that is, that the user knows what a "sticky header"
> is--otherwise, what should it be called?
> "org-header-line-that-shows-the-outline-path-at-the-top-of-the-window"?

I think that org-sticky-header is a fine name.  I think that
"gecko-mode" would be a bad name, even though they are also sticky,
because understanding why the package was given the name is far easier
than inferring the scope of the functionality from the name.

> Now imagine one user trying to share that package with another user:
>
>     Alice: "Hey, I found a package that I think you'll find useful:
>     org-header-line-that-shows-the-outline-path-at-the-top-of-the-window."
>
>     Bob: "Oh, that does sound useful.  What's it called?"
>
>     Alice: "I just told you."
>
>     Bob: "Yes, but what's the name of the package?"
>
>     Alice: "That's it."
>
>     Bob: "No, I need the name so I can install it and try it."
>
>     Alice: "I just told you the name."
>
>     Bob: "No, you told me what it does."
>
>     Alice: "Yes, but that's the name."
>
> (With apologies to Abbott and Costello.)

This is an argument against long names.  The situation I usually have in
mind is when people cannot remember a name, because the name is
arbitrary.  There is a reason why GNOME allows for both
application-specific names and generic names.  I regularly use the
graphical file manager, but don't ask me what the proper name of the
application is.  Something like "nemo" or "nautical"? 

> I'm not merely joking, I'm also serious.  This is a real problem.
>
> It's doubly a problem for a package that has similar functionality to
> existing packages.  For example, with regard to "devil", the package
> recently proposed by Susam Pal, various alternative, ostensibly
> descriptive names have been proposed, like "key-transl",
> "no-modifier-mode", "prefixless-mode", "implicit-ctrl-mode", and
> "comma->control-mode".  If one of those names were used, how would a
> user distinguish such a package from other packages that do similar
> things, such as viper, evil, god-mode, meow, boon, and the numerous
> other similar ones that are out there?  (Note that, although I use
> none of those packages, I remember them because of their distinctive
> names; had they descriptions as names, I would not remember them, nor
> would I be able to easily find them again.)  We're faced with the
> same problem:
>
>     "Hey, I found a useful key-translate package.  You should try it."

The issue here is that you are using an indefinite article.  I think

      Hey, I found a useful package called "key-translate".  You should try it.

is a lot clearer.

>     "Oh, I already use evil-mode."
>
>     "No, not that one.  This one."
>
>     "Well, which one is it?  There are a bunch of those, like viper,
>     evil, god-mode, meow, boon..."
>
>     "key-translate."
>
>     "Yes, I understand that it does that, but what is it called?"
>
>     "The name itself is key-translate."
>
> And that brings us back to the heart of the issue: "Oh, ok.  So how is
> it different than viper, evil, god-mode, meow, or boon?"  That is, the
> user must read the description (or even the whole readme) to
> understand what the package does.  The name is not enough; the name is
> never enough.
>
> And by burdening the name with a responsibility it cannot bear, the
> name suffers, the package suffers, and ultimately, the user suffers.
> The "descriptive" name is not memorable; the user likely forgets what
> it's called a few weeks after installing and configuring it (who
> hasn't experienced this already: installing a package, configuring it,
> and then forgetting about it, finally being unable to remember the
> name of the package that does the thing that one suddenly needs the
> functionality of again, having to search ELPA or MELPA for such a
> tool, and then discovering that one already has it--well, maybe it's
> just me).
>
> Of course, it's a laudable goal to reduce confusion for users.  Yet,
> although Ed is the standard editor, do we not use Emacs?  What do
> potential users say about that?  "Didn't Apple stop making those a
> long time ago?"  Are they not confused?  Should we rename Emacs to
> gnu-lisp-machine-with-built-in-editor?  Would we not shorten such a
> name to GLiMBiE?  And would such a name be descriptive?

There is a counter-tendency to what I mentioned above, and that is that
over time popular names become descriptive in and of itself.  Take
"grep", while we might know what it means and where the name came from,
most people have to memorise the name.  But this ends up being
acceptable, since the term is so widespread.  That being said, I always
remember how when learning how to use a shell, I initially refused to
use "grep", and instead looking for a program like "filter" or "search",
that would have been consistent with the other commands like "mv", "cp",
"cd", etc.  "Grep" sounded like some non-standard utility that I would
have to install from somewhere.

> I'd like to share a link to a short essay published earlier this year
> that reminded me of these threads on emacs-devel:
> https://ntietz.com/blog/name-your-projects-cutesy-things/ Although
> it's mainly about service names rather than software packages, its
> points are relevant here:
>
>     And then the cherry on top, the final nail in the coffin of
>     descriptive names: They're just too hard to say and remember, and
>     they're no fun. I don't want my services or projects to sound like
>     a law firm ("Ingest, Processing, & Storage LLP"). A descriptive
>     name will be wordy, or boring, or both. It won't be memorable, and
>     it won't be fun. On the other hand, something that's cute will be
>     far more memorable and much easier to say.
>
>     The world is boring enough as is. Let's add more whimsy and
>     cuteness through our service and project names.

I think of my father, who despite having started using Unix system in
the 1990's and understands how everything works, just cannot remember
names like "apt" when software has to be upgraded.  That means I have to
do it instead.  One mans joke is another mans arbitrary, opaque
identifier that has to be memorised.

> As an Elisp package author, I can vouch that part of the joy of
> programming is creation, and part of the joy of creation is in naming
> one's creations.  Let us not steal this joy from the authors who
> generously contribute their time and work to the public good that is
> Emacs and ELPA.
>
> I'll close with these words from Alan J. Perlis, quoted in Abelson's
> and Sussmans' classic, /Structure and Interpretation of Computer
> Programs/:
>
>     I think that it's extraordinarily important that we in computer
>     science keep fun in computing. When it started out, it was an
>     awful lot of fun. Of course, the paying customers got shafted
>     every now and then, and after a while we began to take their
>     complaints seriously. We began to feel as if we really were
>     responsible for the successful, error-free perfect use of these
>     machines. I don't think we are. I think we're responsible for
>     stretching them, setting them off in new directions, and keeping
>     fun in the house. I hope the field of computer science never loses
>     its sense of fun. Above all, I hope we don't become
>     missionaries. Don't feel as if you're Bible salesmen. The world
>     has too many of those already. What you know about computing other
>     people will learn. Don't feel as if the key to successful
>     computing is only in your hands. What's in your hands, I think and
>     hope, is intelligence: the ability to see the machine as more than
>     when you were first led up to it, that you can make it more.

Re: Names are not descriptions; descriptions are not names

[Eli Zaretskii]

> Date: Thu, 11 May 2023 23:01:24 -0500
> From: Adam Porter <adam@alphapapa.net>
> 
> There is no better way to doom software to obscurity than to give it a
> description as its name.  Names are not descriptions; descriptions are
> not names.

No one said names should be descriptions.  The request was to make the
name of a package include some hint on what the package does and where
it is applicable.  Just a hint, nothing more.

> It's doubly a problem for a package that has similar functionality to
> existing packages.  For example, with regard to "devil", the package
> recently proposed by Susam Pal, various alternative, ostensibly
> descriptive names have been proposed, like "key-transl",
> "no-modifier-mode", "prefixless-mode", "implicit-ctrl-mode", and
> "comma->control-mode".  If one of those names were used, how would a
> user distinguish such a package from other packages that do similar
> things, such as viper, evil, god-mode, meow, boon, and the numerous
> other similar ones that are out there?

Again, there's no requirement to allow users distinguishing between
similar packages just by looking at the name.  That can only be done
by reading the package's description.

IOW, the name should allow some kind of initial filtering: if I'm not
interested in key translation, I don't need to look further at a
package named key-transl.  It's the same first-order filtering we
apply to email by just looking at the Subject.  Nothing more, nothing
less.  We advise people to use descriptive enough Subject in their
email messages so that interested people will take notice; sending a
message whose Subject says "42" will only catch attention of a group
of people who have read the book and remember the significance of the
number, but not others.  Likewise with package names.

> And by burdening the name with a responsibility it cannot bear, the
> name suffers, the package suffers, and ultimately, the user suffers.
> The "descriptive" name is not memorable; the user likely forgets what
> it's called a few weeks after installing and configuring it (who
> hasn't experienced this already: installing a package, configuring it,
> and then forgetting about it, finally being unable to remember the
> name of the package that does the thing that one suddenly needs the
> functionality of again, having to search ELPA or MELPA for such a
> tool, and then discovering that one already has it--well, maybe it's
> just me).

How many package names can a user remember? 10? 20?

There are many hundreds of packages out there.  No one can possibly
memorize them, even if each one of them had a catchy name.  It is
impractical to expect that, and therefore requesting each package to
have a catchy name is not very important: its goal cannot be reached
in practice anyway, so why request that?

> As an Elisp package author, I can vouch that part of the joy of
> programming is creation, and part of the joy of creation is in naming
> one's creations.  Let us not steal this joy from the authors who
> generously contribute their time and work to the public good that is
> Emacs and ELPA.

I see your point, but please also see ours.  We are not looking at
this via a loophole of an author of several packages, we are looking
at this from the wider POV of managing Emacs as a project.  From our
POV, having packages whose names don't even hint on their purpose is a
significant disadvantage.  So we respectfully request package authors
to cooperate by coming up with names which don't have this downside.
If package authors can come up with smart names that also hint on
their use areas, the joy you mention can be preserved, but without the
downsides.

Re: Names are not descriptions; descriptions are not names

[Jim Porter]

On 5/12/2023 12:37 AM, Eli Zaretskii wrote:
> Again, there's no requirement to allow users distinguishing between
> similar packages just by looking at the name.  That can only be done
> by reading the package's description.
> 
> IOW, the name should allow some kind of initial filtering: if I'm not
> interested in key translation, I don't need to look further at a
> package named key-transl.  It's the same first-order filtering we
> apply to email by just looking at the Subject.  Nothing more, nothing
> less.

Looking at the list of packages in GNU ELPA, almost all of them have a 
description well within the limits of what we'd expect from an email 
Subject. To use your example of "42", while that alone would be cryptic, 
a Subject like "42: The answer to life, the universe, and everything" 
would be ok, I think.

That being said, are there places we can make package descriptions more 
visible? Both "M-x list-packages" and the ELPA package list on the web 
show the descriptions prominently, so I find them both to be very easy 
to find packages, even if they have cryptic names. However, there may be 
further improvements we should make in this area. Packages can have 
keywords, so improving those might help, or maybe we could add metadata 
for broad package categories (e.g. "theme", "key bindings", "major 
mode", etc)...

Still, there's room for a light touch with improving package names 
(especially if we let the actual package *identifier* be a slight 
variation on the name). For example, the Devil package could otherwise 
stay the same, but we could give it a package identifier of 
"devil-keys". The functions and commands would still just be 
"devil-FOO", and the documentation could still say "devil" instead of 
"devil-keys" (except when talking about how to install the package, of 
course). Would that be a reasonable compromise for cases like this?