|
From: Alan M. <a_j...@ya...> - 2007-01-14 12:22:49
|
John, Thanks I will start looking at the code when I get a chance. Below are some spontaneous remarks.... > I put a makefile in the root so my :make will still > produce the desired > results, and added a quick distribution method to > the build.xml so I can > easily get updated javadocs onto our web site. I presume that we would use ant only - do we really to bother with make (aside from anything else you might want to use it for) ? > Alan, > Take a look at it and see if you can figure out what > I was thinking, [snip] I'll play around with it can ask the relevant questions later ;) > I have not even looking into adding JUnit support > but realize a test suite > would be of great value for this thing as we chug > away at it. Junit good, in this case probably essential as we (and end-users) need to be sure of quality and test coverage. If you like, consider writing the unit tests first and do the implementation later, that way you get an early idea of how idiomatic the library code will be... > 1. I am absolutely addicted to the Jakarta commons > and I'm even considering > naming a child Jakarta, so we are using the > HttpClient library from that > Apache project, it was easy and already in other > parts of my application at > the time. However, this should not become a weight > around our necks, it is > not so difficult to use the standard java > URLConnection handlers to pull the > required data back, so if someone comes along and > decides those guys are > communists and the implementation should not be > dependent on that, let them > enjoy freeing us from the dependency. Have you considered using a facade and hidding the HttpClient stuff behind it (that way it can be replaced easier later if no longer required?) - this might be a secondary consideration though because the concept still requires proving.... > 2. I also chose to use log4j for logging. I really > like log4j, [snip] log4j also good! Some Apache projects prefer (and sometimes irritatingly enforce) the use of another logger e.g., commons-logging etc. but I still think we should stick with log4j. > 3. the logic for applying filters and processors may > need some review, in a > few places I wasn't sure which method to call so I > tended to call the method > I had the most data available to send, which may > prove shallow and > unpredictable during implementations, I will try to > organize a way to > articulate the logic as it exists today. Perhaps design first (e.g., UML) and implement later ? That way we can see what you are thinking (literally!) BTW: I use the Umbrello UML editor > And that is my typing quota for the day, good luck > and thanks again for > joining this effort, may you all become immortal for > your efforts, women > will flock to you, children will sing songs of > praise, Porsche will name a > car after you, and one day you will evolve into pure > energy and come to know > the true nature of God. Or something like that :) ...failing that women will flee us, children will cry in our presence...no wait, that's what happens at the moment. ok things can only get better ;) atb, A. --------------------------------------------------- You can find my GPG key (a_j...@ya...) at ETH PGP Keyserver: http://www.tik.ee.ethz.ch/~pgp/ --------------------------------------------------- ___________________________________________________________ Der frühe Vogel fängt den Wurm. Hier gelangen Sie zum neuen Yahoo! Mail: http://mail.yahoo.de |