For testing during development, you don't need to do what you want. Certainly, for testing before you are done and before you turn it in, test it with a lot of real world stuff. For that you can get creative, like use current and previous versions of any distro you want, diff them independently of your torernt-doohickey, transport then files with your doohickey and then independently verify that the diff on the receiving end is identical to the diff on the sending end. But for development, you should probably consider that you want a small package to test with so you can reduce your edit-compile-test cycle. A diff is a diff is a diff. It doesn't matter if you are diffing a 700MB linux distro, or a 1MB tarball. Get the algorithm right, and size becomes a matter of scaling, not a hindrance to your development. The point afterall is to create differential torrents, not to create linux isos.
You should maximize your test set for ease of development without compromising the validity of the tests. When you are done with development, go nuts on testing it with all sorts of stuff. After all, almost every other engineering discipline in the world starts with small models of the real thing for testing, before moving on to the real thing. It only makes sense to apply the same rules to software engineering. But hey, do what you want. -Michael ----- Original Message ---- From: Brian Hawkins <[EMAIL PROTECTED]> To: Provo Linux Users Group Mailing List <[email protected]> Sent: Thursday, March 8, 2007 2:24:54 PM Subject: Re: Create my own linux ISO Yes that and I want a real life data set. I'm trying for something that is commonly downloaded using bittorrent. Linux ISO images are what came to mind first. I think I can get done what I need by tearing apart an image and them putting it back together with different compression techniques. Thanks Brian Nicholas Leippe wrote: > On Thursday 08 March 2007 12:16, Michael Brailsford wrote: > >> Why not do something like create a directory called /tmp/test_data and copy >> in a few header files or other text files, maybe copy in the whole >> /usr/include directory to get a bigger size. Then you can do: >> >> $ tar czf /tmp/test.iso /tmp/test_data/* >> >> Then you can just make an edit in a file in /tmp/test_data and recreate >> your "iso", then transfer it and compare what you get on the other end with >> what you sent. >> >> A whole linux distro iso, while nifty, is significant overkill. >> > > It may be overkill for verifying the correctness of his implementation, but I > think he wants a bit larger data set for better performance metrics. > > > /* > PLUG: http://plug.org, #utah on irc.freenode.net > Unsubscribe: http://plug.org/mailman/options/plug > Don't fear the penguin. > */ > > /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */ /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
