resurrect libmagic

[buhman]

20:55:10 sudokode haha @ loading 'tQ6j'
20:55:28 sudokode nice png of yosemite. the textarea handles that binary quite nicely :D

20:59:07 buhman I've considered re-adding libmagic where all text/* is overridden to text/plain
20:59:17 buhman which would simulate libdumb behavior I think                                  
21:00:01 Earnestly buhman: Pretty much

[buhman]

85c6fba2aadca7687629e001b4d694aa4ec9617d + 829a40be6e8cd9cb07eab86a455f3aaf82d5e683

[buhman]

worst idea ever; 8046dcb76d6189085d2e328ac6625147309e1106

[anarcat]

worst idea ever; 8046dcb

why, out of curiosity, was this approach abandoned? i understand we can just add a suffix to guess the mimetype, but it would be great to have automation on the server-side, or at least a way for clients to provide a mimetype... :) thanks!

[buhman]

This was over 2 years ago; I know about as much as these commits as you do :)

or at least a way for clients to provide a mimetype

There is, actually: https://ptpb.pw/a#header-paste-media-types

Here's how you do it in curl:

λ zack [~] → curl -F 'c=@-;filename=_.png' https://ptpb.pw < foo.jpg            
date: 2018-08-25T18:01:14.059523+00:00
digest: 14607bd8db563d47e3ffa57dba0ea0fcd3a167b3
long: ABRge9jbVj1H4_-lfboOoPzToWez
short: oWez
size: 631936
status: created
url: https://ptpb.pw/oWez.png
uuid: 3f6d9dea-6aeb-47c6-9d50-6655c4621759
$ curl -sD- https://ptpb.pw/oWez | grep Content-Type
Content-Type: image/png

By default filename is the same as the form name value (in this example, -). IOW if you did:

$ curl -F 'c=foo.jpg' https://ptpb.pw
$ curl -sD- https://ptpb.pw/oWez | grep Content-Type
Content-Type: image/jpeg

..the filename behavior is implicit. If filename can be evaluated to a mimetype known to pb, that mimetype gets stored in the paste metadata at creation.

[buhman]

I never actually do any of this, instead I use the per-request (${id}.png) behavior; it's marginally less work than filename twiddling and better hints to the recipient what is about to be loaded by their browser.

[anarcat]

the thing is i'm looking at integration into other pastebin clients (e.g. pastebinit) and one thing they can do is send a mimetype along, so it would be useful to document those things better in the API or somewhere... :)

thanks!

[buhman]

Oh, that's a little different from what this issue was dealing with (magically calculating mimetype by examining the paste body).

pastebinit

I don't understand why a 500 line script is considered desirable for this, especially one so awkwardly written (imports and function definitions wrapped in try/except; NIH'ing six compatibility imports; awkward "params" building).

Anyway, if you want to use this, maybe try json (the alternative is urlencoded, which works only by accident, not as a thing I'm interested in supporting).

This should be something like:

[format]
content = content

[defaults]
post_format = json

[anarcat]

eh. i don't like pastebinit much either, but it does have the advantage of talking with many pastebins and providing a unified interface... :) and really, so far I've just used your simple shell script (or just remember the curl magic) instead and i love it...

i will definitely try your suggestion next time i try to fix pastebinit again though, thanks!