At 02:46 16.03.2001 +1100, Peter Donald wrote:
>At 04:17  15/3/01 +0100, Ceki Gülcü wrote:
>>>A lot of projects use dist/, in my personal projects I use out/ while some
>>>projects use bin/. 
>>
>>See, these names are like colors, blue, green, or yellow... At face value,
>they are all equivalent. That's what I meant by sticking to a name. It goes
>without saying that the chosen color has to be reasonable, unlike #A26F58
>or SemiHuedVelvetBlue.
>
>True but we tend to have some ... err .. strong willed individuals here who
>we will find difficult to move from their pet structure. You can make a
>standard all you want but we have to get people to follow it ;) 

I hear you. 

>>Forgive my ignorance but what is so wrong with putting "raw" class files in 
>>the distrib along jar files? Is it the wasted disk space?
>
>messy classpath, wasted diskspace, incredibly slow access on certain
>filesystems (ie FAT) etc. Also much more likely to break the 100 char limit
>in tar files etc.

One is free to use jar files, using the class/ dir is an option. The tar char limit is 
broken by a class file then it is likely to be broken by that file's javadoc 
equivalent as well. How serious is this tar limitation anyway?

>>I am not aware of the assumption that the dist directory is meant the
>unzipped 
>>version of the binary distribution. Not at all. 
>
>well theres the rub.
>
>>
>>If the files
>>
>>xdocs/manual.xml
>>build/build.xml
>>build/build.sh
>>build/build.bat
>>build/lib/ant.jar
>>build/lib/xerces.jar
>>src/x.java  
>>
>>are contained in the CVS module for project X, then after compiling,
>generating javadocs, jarring, and building a distribution image, one would
>get:
>>
>>docs/manual.html
>>docs/api/
>>docs/api/x.html
>>docs/api/overview.html
>>docs/api/index.html
>>docs/api/etc...
>>build/build.xml
>>build/build.sh
>>build/build.bat
>>build/lib/ant.jar
>>build/lib/xerces.jar
>>src/x.java  
>>xdocs/manual.xml
>>dist/classes/x.class
>>dist/lib/X.jar
>>
>>dist/images/X.tar.gz
>>dist/images/X.zip
>
>Maybe in *your* model ... 

Yes, that's the model log4j uses although I am adapting the log4j directory structure 
to match the dirlayout document and this ongoing discussion.

>>where X.tar.gz and X.zip contain all the above files other than
>themselves. The images can be refined to contain only source code,
>documentation, or binaries. Does that make sense? Ceki 
>
>It's one way of doing it but it requires developer sentience whilst
>building distributions - something I try to avoid at all costs ;) Besides
>that model only works for certain types of projects. Some projects build
>particular types of images etc. Having everything below dist/* in binary
>distribution is much easier and is a common practice that I believe
>originated in java.apace.org projects and migrated to jakarta/xml groups.


OK but I don't see any big difference between the two approaches. In one you generate 
into dist/ + copy into dist/ and the other you tar/zip a set of files specified in a 
tar/zip target.  You have to be sentient in both cases or? Ceki 



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to