bug#22567: Factoring 38 nines

[Leslie S Satenstein]

Works for me, but it took a long time  
 Regards 
 Leslie
 Mr. Leslie Satenstein
Montréal Québec, Canada



 
      From: Eric Blake <address@hidden>
 To: SasQ <address@hidden>; address@hidden 
 Sent: Friday, February 5, 2016 2:29 PM
 Subject: bug#22567: Factoring 38 nines
   
On 02/05/2016 11:30 AM, SasQ wrote:

> 
> OK, this convinces me this is not a bug. 4m30 on my machine.
> But it's definitely a user-interface fail ;)
> It should at least output some warning that the computations might
> take longer, or display some progress status / estimated time along
> the way.

Estimated status is very hard to produce. I guess with trial division,
you could output a progress indicator comparing what number you are
trying in relation to the square root of the number being factored, as a
rough estimate; but there's still the annoying problem that any estimate
bar will not progress smoothly (once a factor is found, the time
remaining changes).  Progress indications should not be issued by
default, unfortunately, because it might break expectations of
applications that have grown used to no progress, so you'd have to make
it a new command line option - but then, how will people know to use the
new option?

> Because otherwise the user can think it simply hangs.

If a user is naive enough to think it is hanging on large input, how can
we expect them to also be aware that they can turn on an option to track
progress? And how will we explain that the progress meter may have no
bearing on the real amount of time required?

> 
>> The source code is there for you to peruse.
> 
> There sure is, but analyzing it just to figure out the algorithm takes
> much more time than refering the maual to see which particular
> factorization algorithm or its variation is in use.
> 
> It took me a while to find the answer on StackOverflow:
> http://stackoverflow.com/questions/11155331/what-is-the-algorithm-behind-the-factor-command-in-linux
> so mentioning it in the man page wouldn't hurt.

Well, it IS mentioned in the documentation (the info page, as the
preferred documentation for a GNU project); and the point of the man
page is to be a terse summary, not the full documentation.  But maybe
you have a valid point that adding a one-line blurb mentioning the
Pollar Rho algorithm in 'factor --help' (which in turn feeds the man
page) might be a useful change.

-- 
Eric Blake  eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org