[IPOL discuss] The Hatchery: an informal index of available code within our community.

[Pierre Moulon]

Hi,

I want to give alarm about sharing code.
Since that C++ exist there is no standard as STL in order to manage IMAGE
I/O and pixel manipulation ...

Sharing code is a nice idea, but in order to use code made by another people
is not so easy.
You do not know the normal use and which limits the function code handle...
   code must be commented... and/or unit tested...

Every code must come with an example usage that show the basic usage in
order to learn the user how
to use the piece of code. A unit test is even better.

About using external library...
cons:
  - it makes the algorithm often library dependent...
  - it makes your algorithm dependent of the library's potential bug,
limitation of the library.
pros:
 - it makes often the code more readable trough the usage of high level
interface.
 - you write and suppose that the code you use is bug free.


So I think Ipol submission code could be seen in a two part code :
 - A generic code (writing, reading, drawing image code, handle command line
parameter)
 - A submission code (the core of the submitted algorithm) that must use
very simple High Level interface.
   - Using high level interface make the reading of the algorithm easier to
the reviewer,
      to the user...
     and makes "face to face" with the pseudo code on the Ipol webpage
easier.

So I think the Hatchery must be compound of *tiny* independent library*.*
*Make it tiny, make maintenance easier. *and it do not create hard
dependence over a given library !*
*
*
*
*A potential drawback could be in the future to force IPOL submitter to
use/recommend this Hatchery...
*
*-> We must focus on algorithm and exact interpretation of algorithm, not on
software libraries.
*
*
*
*I will finish my mail with my "adage"
*
*"If you try to make a wheel by yourself you could make it not really
round... But, using a existing wheel you could invent a Bike, a Car..."
*
*But you have to check each elements connection to be sure of the results
;-) ... (Quality checking .... is another story...)
*
*
*
Cheers,
Pierre
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tools.ipol.im/mailman/archive/discuss/attachments/20110927/4dcad88f/attachment.html>