Re: DB_MetaReport - Metadata reporting libraries for PEAR

Mark Cotner wrote:

Please do bear in mind that, in order of priority, ease of use, performance, and being lightweight were the fundamental goals. I also tried to limit dependencies to make installation easier. When I reviewed packages for use like HTML_Table and Pager it seemed much faster/easier to output my own instead of including a fairly large toolkit. Eventually I can make them optional and only include if specified by the user.

Yep dependencies are a tricky topic. PEAR doesnt encourage developers to go all out on dependencies. For examnple just because you open a file, doesnt mean you need to include the PEAR file package. Anyways its a call the original author has to make. However there are severe advantages to dependencies if done right. For example I doubt that your pager handling is as flexible as the Pager package. It also probably hasnt received the amount of attention to detail as the Pager class has. So to me if you have to do more than .. lets say 10 lines of code .. on something then you may want to consider using a dependency instead. If you start writing multiple methods to handle the same then you are probably better off with a dependency. One thing that we have been encouraging though is to provide "lite" versions of packages that follow the parent package in API but only provide a limited set of features. So before you come up or maintain your own variation of something because the package in PEAR feels too heavy, then exploring to come up with a "lite" package might be another option. This will enable people to scale up to the "real thing". Provides you with an API that has already some thought behind it etc. regards, Lukas