Re: [tlug] rsync efficiency (was: The Mother of All (bash) Commands)

["Josh Glover" <jmglov@example.com>]

On 01/04/2008, Attila Kinali <attila@example.com> wrote:

> Yes, but would you call a sorting algorithm fast, if it would
>  perform good on random sorted data (O(n*log(n)), but take o(exp(n))
>  if the data is reverse ordered? Yes, it might be a pathological
>  case, but if this pathological case is actually quite common,
>  then the algorithm as a big flaw.

The point is not whether you would call the algorithm fast or slow,
the point is knowing when to use it and when not to. For example,
bubble sort is an excellent algorithm if I know my data sets are
guaranteed to be small, since it is easy to implement compared to, say
quicksort or mergesort.

So if you know you have a pathological data set, choose a different
algorithm, but don't say that the algorithm is worthless when it
performs quite well on data sets that correspond to its design goals.

Remember, before rsync, people would always use tar + scp (well, tar +
uucp, I guess). Rsync was designed for the situation where you have
relatively large files (i.e. not Maildir trees!) with relatively small
changes; if you are in that situation, use rsync. If not, stick with
some combination of tar / dump and scp / ssh / nc.

-- 
Cheers,
Josh