Re: length of a package extended description

[Ian Jackson <ijackson@chiark.greenend.org.uk>]

Adam Borowski writes ("Re: length of a package extended description"):
> Some data: count of packages with descs of a given length:
...

Here's Adam's data with cumulative package count, and cumulative
percentage:

>   1-  4 13772 13772  30%
>   5-  9 21324 35096  77%
>  10- 14 6531  41627  91%
>  15- 19 2433  44060  96%
>  20- 24 872   44932  98%
>  25- 29 288   45220  99%
>  30- 39 175   45395  99%
>  40- 49 42    45437  99%
>  50- 59 19    45456  99%
>  60- 69 13    45469  99%
>  70- 79 5     45474  99%
>  80- 89 7     45481  99%
>  90- 89 3     45484  99%
> 100     1     45485  99%
> 110-119 7     45492  99%
> 120-150 3     45495  99%
> 151-199 4     45499  99%
> 203     1     45500  99%
> 257     1     45501  99%
> 277     1     45502  99%
> 325     1     45503  99%
> 350     1     45504  99%
> 437     1     45505  99%
> 1935    1     45506  100%

Of course as maintainers we all have a natural tendency to think our
own package is more important and interesting than other packages.
That kind of comes with the territory.  That means that the average
description will tend to be longer than any agreed target average
description length.

But, worse, it appears that some maintainers aren't able to exercise
their discretion on this question in a manner which most of the rest
of us would consider reasonable.

IMO a proportionate response would be a target, and a hard limit, in
policy.

I would say:

  The extended Description should ideally fit within 14 lines.
  It must not be longer than 24 lines.

The target of 14 lines would be missed only by 8.5% of packages and
the limit of 24 lines breached only by 1.26% of packages.

Ian.