Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[ssd]

* FRIGN 2015-12-12 20:45
> The fun thing is: sent currently supports pbm as well when you run 2ff, 
> because
> it internally calls imagemagick's "convert" in case it detects a file format
> not handled by the native farbfeld tools.
> This means that if you supply sent with a pbm file, it will run it through
> 2ff just fine. Try it yourself. :)

ironically, I don't have imagemagick installed on any of my boxes, but
have netpbm on all. (but then I can use any of the numerous ppmto* and
pnmto* and pamto* to convert to whatever I need.)

that said, however, I don't really get why one would insist on having
pbm as intermediate format, apart from "because you can". 

for me, pbm is a medium which allows to apply all the tools available in
the netpbm suite on a bulk of images, and there are quite some of them.

cheers
--s

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[David Phillips]

On Sun, Dec 13, 2015 at 06:40:27AM +0100, FRIGN wrote:
> On Sun, 13 Dec 2015 11:26:34 +1300
> dbphillip...@gmail.com wrote:
> 
> > Valid observation, although this patch is about changing the intermediate
> > format, not the input format.
> 
> I know, but what argument can you give? I mean, this entire thing is only
> based on spread- and team-arguments.

I won't personally give an argument, I feel like farbfeld is nice and
standard and fits this purpose a little better than netpbm.

>From my perspective, netpbm has too many extensions and "flavours" to
provide support without an external library.

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[FRIGN]

On Sun, 13 Dec 2015 11:22:30 +0100
s...@mailless.org wrote:

> ironically, I don't have imagemagick installed on any of my boxes, but
> have netpbm on all. (but then I can use any of the numerous ppmto* and
> pnmto* and pamto* to convert to whatever I need.)

I don't blame you, imagemagick is a beast. But if you do graphics stuff,
the pdf- and other utilites are very handy.

> that said, however, I don't really get why one would insist on having
> pbm as intermediate format, apart from "because you can". 

Exactly.

> for me, pbm is a medium which allows to apply all the tools available in
> the netpbm suite on a bulk of images, and there are quite some of them.

I'll think about adding pbm2ff and ff2pbm to the farbfeld-utilities :P
Then the crowd is happy and you can use the netpbm suite.

Cheers

FRIGN

-- 
FRIGN 

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[FRIGN]

On Sun, 13 Dec 2015 11:26:34 +1300
dbphillip...@gmail.com wrote:

> Valid observation, although this patch is about changing the intermediate
> format, not the input format.

I know, but what argument can you give? I mean, this entire thing is only
based on spread- and team-arguments.
The technical side has been lacking for quite a while now.

I understand, netpbm is _the_ format here and I shouldn't dare to question
the fact that it is widespread as hell.
But apart from widespread, what is a reason to use it? The library interface
is as cumbersome to use as libpng. There are good reasons why sent doesn't
use PNG as an intermediate format.

So, what is the reason? What does netpbm really _do_ better than farbfeld?
And no, Roberto, writing an image in ed does not cut it.

Cheers

FRIGN

-- 
FRIGN 

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[dbphillipsnz]

On Sat, Dec 12, 2015 at 08:45:49PM +0100, FRIGN wrote:
> On Fri, 11 Dec 2015 11:59:33 +0100
> Markus Teich  wrote:
> 
> > no it will not, but as noted in the other mail you are welcome to push your
> > patch to the wiki.
> 
> The fun thing is: sent currently supports pbm as well when you run 2ff, 
> because
> it internally calls imagemagick's "convert" in case it detects a file format
> not handled by the native farbfeld tools.

Valid observation, although this patch is about changing the intermediate
format, not the input format.

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Markus Teich]

Grant Mathews wrote:
> Aside from those points: would rewriting the patch to not use libnetpbm,
> but to use the Netpbm helpers instead, be acceptable?

Heyho Grant,

no it will not, but as noted in the other mail you are welcome to push your
patch to the wiki.

--Markus

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Markus Teich]

Heyho Grant,

Grant Mathews wrote:
> On Thu, Dec 10, 2015 at 04:39:49PM +0100, FRIGN wrote:
> > Netpbm is arguably almost more complex than BMP and not easy to handle.
> 
> It's literally the top result when I search for the phrase "simplest image
> format".

So your world is limited by the first ten pages of a search engine?

> > We could discuss it if it was widely used,
> 
> Again, it's the top result for simplest image format: people use it.
> dcraw uses it, it's supported by every image viewer I've ever installed,
> and it shows up in enough random places that I'd consider it pretty
> standard.
> 
> farbfeld is used nowhere, and sets the bar for "widely used" quite low.

It's obvious, that pnm is used by more people than farbfeld atm, however that
doesn't make it better. farbfeld was just announced a few weeks ago. If we would
just go by widespread usage, we probably would all write crappy shit in C++ with
boost.

> > but the main point is: __You need a library to handle it__ and it's
> > not that much of a popular format to justify installing a library for
> > it.
> 
> You don't actually need a library for it, I'm just really lazy and the
> library handles all the hard parts for me.
> 
> If the library "requirement" is the only sticking point, it'd be pretty
> easy to fix that.

No it's not, but it was the main motivation for dropping libpng in the first
place and we had quite a bit of discussion for the change on the slcon and the
ML as well.

> > I don't know about you guys, but I don't have libpbm installed on my
> > computer and even though ffmpeg for instance offers support, I might
> > be having a hard time finding a format ffmpeg _doesn't_ support.
> 
> I don't have farbfeld installed on my computer, and ffmpeg doesn't even
> support it.

Right, then install it, it's easy! If you're running gentoo and lazy, I can even
provide you with an ebuild from my overlay.

> > A question for the diligent reader: Can you read in a netpbm file without
> > first looking into the docs?
> 
> I, uh, what? Are you proposing a format that doesn't require explanation?
> Without any sort of docs (or reverse engineering), *no* format is readable.
> Though, to be fair, PAM comes close: the header is pure ASCII, and fairly
> self-explanatory.

Let's put it that way: I just had a look at the user manual[0] and after reading
about the same length as the farbfeld README and FORMAT files, I noticed I don't
have time to read the full doc and the only useful thing I learned was that
there are different formats for color, grayscale and b/w images, so netpbm seems
to redundantly do the job compression algorithms were written for.

> Really, I just want this tool to be usable. Requiring some ridiculously
> obscure image format for no reason takes this from "something I might
> want to use" to "something nobody is going to use".

Well nobody is forcing you to use it. You're also welcome to put your unaccepted
patch on the wiki. Maybe there are other fans of netpbm who would want to use
it? Have you read the README and FORMAT files from farbfeld? It only takes 5
minutes.

--Markus


0: http://netpbm.sourceforge.net/doc/

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Matthew of Boswell]

On Fri, 11 Dec 2015 11:59:33 +0100
Markus Teich  wrote:

> Grant Mathews wrote:
> > Aside from those points: would rewriting the patch to not use libnetpbm,
> > but to use the Netpbm helpers instead, be acceptable?  
> 
> Heyho Grant,
> 
> no it will not, but as noted in the other mail you are welcome to push your
> patch to the wiki.
> 
> --Markus

It seems a few people wanted pbm support instead of farfeld in sent, so
it's probably a good candidate for the wiki.

Personally, I don't see much reason to have a complex image format as
an intermediate when all we really care about in sent is simplicity
and the internal representation of image data. Farbfeld is the magic
"read any type of image" component, and it shouldn't matter whether you
use farbfeld or some magic bloated library that imports all image
formats.

If you replace the word "farbfeld" with "libimagereader", would you
still be motivated to replace it with pbm? Maybe calling farbfeld an
image format (technically correct) is throwing people off. "farbfeld
the image file reader"? But I don't think that description would portray
the special simplicity of farbfeld.

-- 
Matt Boswell

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Markus Teich]

FRIGN wrote:
> __You need a library to handle it__ and it's not that much of a popular format
> to justify installing a library for it.

Heyho,

like FRIGN pointed out correctly this patch would negate the effort and goal of
the switch to farbfeld. sent with farbfeld uses a separate process instead of
linking to a library and is therefore more UNIXy and sucks less. The patch won't
be merged.

--Markus

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[FRIGN]

On Wed, 9 Dec 2015 20:56:38 -0800
Grant Mathews  wrote:

Hey Grant,

> Since the PNM/PAM format already exist as a minimal intermediate
> representation with a rich set of commandline tools to manipulate them,
> use Netpbm to handle images.

back to your desk, Grant, you've missed the purpose of farbfeld.

No, to be fair, we discussed the PAM-format at slcon2 in more than one
coffee break.
Netpbm is arguably almost more complex than BMP and not easy to handle.
We could discuss it if it was widely used, but the main point is:
__You need a library to handle it__ and it's not that much of a popular
format to justify installing a library for it. I don't know about you
guys, but I don't have libpbm installed on my computer and even though
ffmpeg for instance offers support, I might be having a hard time
finding a format ffmpeg _doesn't_ support.

You've replaced the entire farbfeld parsing code in sent with the
boilerplate "offered" by netpbm. Good job!

A question for the diligent reader: Can you read in a netpbm file
without first looking into the docs?

Cheers

FRIGN

-- 
FRIGN 

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Grant Mathews]

On Fri, Dec 11, 2015 at 01:30:52AM +0100, Markus Teich wrote:
> sent with farbfeld uses a separate process instead of linking to a
> library and is therefore more UNIXy and sucks less.

This patch still does all the file conversion in a separate process, and
sent still uses an external library regardless of whether this patch is
merged.

I'm a big fan of process-level code reuse, but no tool is right for
every job, and sometimes a library is the correct way to go.

Aside from those points: would rewriting the patch to not use libnetpbm,
but to use the Netpbm helpers instead, be acceptable?

- Grant

Re: [hackers] [sent] [PATCH 2/2] replace farbfeld with libnetpbm

[Grant Mathews]

On Thu, Dec 10, 2015 at 04:39:49PM +0100, FRIGN wrote:
> Netpbm is arguably almost more complex than BMP and not easy to handle.

It's literally the top result when I search for the phrase "simplest
image format". I've written PPM loaders by hand in projects too small
for a makefile.

> We could discuss it if it was widely used,

Again, it's the top result for simplest image format: people use it.
dcraw uses it, it's supported by every image viewer I've ever installed,
and it shows up in enough random places that I'd consider it pretty
standard.

farbfeld is used nowhere, and sets the bar for "widely used" quite low.

> but the main point is: __You need a library to handle it__ and it's
> not that much of a popular format to justify installing a library for
> it.

You don't actually need a library for it, I'm just really lazy and the
library handles all the hard parts for me.

If the library "requirement" is the only sticking point, it'd be pretty
easy to fix that.

> I don't know about you guys, but I don't have libpbm installed on my
> computer and even though ffmpeg for instance offers support, I might
> be having a hard time finding a format ffmpeg _doesn't_ support.

I don't have farbfeld installed on my computer, and ffmpeg doesn't even
support it.

> You've replaced the entire farbfeld parsing code in sent with the
> boilerplate "offered" by netpbm. Good job!

Yes, that is not a bad thing.

> A question for the diligent reader: Can you read in a netpbm file
> without first looking into the docs?

I, uh, what? Are you proposing a format that doesn't require
explanation? Without any sort of docs (or reverse engineering), *no*
format is readable. Though, to be fair, PAM comes close: the header is
pure ASCII, and fairly self-explanatory.


Really, I just want this tool to be usable. Requiring some ridiculously
obscure image format for no reason takes this from "something I might
want to use" to "something nobody is going to use".

- Grant