Re: CVS commit: pkgsrc

[John Marino <dragonflybsd%marino.st@localhost>]

On 4/10/2013 19:17, Aleksej Saushev wrote:
> John Marino<netbsd%marino.st@localhost>  writes:
> The number "24" is exactly the arbitrary one.
>
> First, the standard for as long as the recent decade is having scrolling
> capabilities even in the modest consoles modes.


Having the ability to scroll doesn't justify the existence unnecessarily 
verbose descriptions.


> Second, were you more attentive to what actually happens than to pkglint
> warnings you would note that with "no-scroll" requirement DESCR should
> be much shorter because pkg_info output takes about twelve lines to
> print package name, short comment, package requirements, project home
> page, and so on. Thus the number "24" has no relation to the problem
> of fitting pkg_info output to the number of lines on archaic console.


That's a good point.  I will happily support a shorter limit.  However, 
I was talking about "> cat DESCR" so my 24-line comment is perfectly 
valid.  We're talking about command-line use, remember?  pkg_info 
doesn't help for packages that aren't yet installed.


> Third, if you personally don't like scrolling, you can use "pkg_info -c",
> and thus DESCR is not meant for you. It is mostly for those users, who
> browse generated HTML index on site or locally, or those who want to
> understand why there're so many p5-Something-Something-Something
> packages and what role they serve without guessing from concise comment.


I disagree.  Greg Troxel already addressed this well.  There's no 
justification for anything longer than the equivalent of an abstract. 
It doesn't matter if it appears in a browser or xterm.


> > The original packager failed to follow the these guidelines, so
> > fixing it now (however it's done) is legitimate.
>
> Truncating them arbitrarily doesn't help anything.


There is nothing "arbitrary" about creating a concise summary.  I 
realize that the description is in English and that for many developers 
English is not their native language.  However, a 10 year-old has the 
ability to summarize a topic in a few sentences.  I didn't see the 
commits, but rodent indicated his editing was not arbitrary.  "kept most 
info" implies an intentional process.


> How much time do you spend reading pkg_info output in archaic console mode?
> How does that apply to other users?


1. I don't use pkg_info, I use (more|less|cat).  2, I spend all my time 
in console mode.  3, console mode is lowest common denominator, it's 
reasonable to assume this is what is targeted, not a browser.


> If you want pkglint to be useful tool, fix it then.


Part of your issue is that you don't respect standards.  For example, 
technically there is nothing wrong with the unit of inches.  If you 
produced a blueprint in inches, and a scanner flagged it and said, 
"imperial units detected, switch to SI units", you'd respond the scanner 
was stupid and imperial units are perfectly legitimate to use.  Well, 
that's true, but you aren't respecting a declared standard.   So I can 
project that it doesn't matter how improved the pkglint tool is made, 
you probably wouldn't pay attention to it anyway.  Am I way off?