hi - just getting started w/ hadoop. great stuff.
firstly, can we consider the following nit patch, to address: ubuntu 6.10b uses dash for /bin/sh vs bash and as such the "==" operator is not supported this mod should work fine for sh, bash and dash append ".txt" suffix to the log files so that the the contents can be displayed via the admin screens now, re dfs ... i'd like to prototype some work around dfs regarding classic file dispurtion, fragmentation, querying and reconstituation. before i deep dive on my own, i though i'd solicit some input from the good people of this list. i've got 2 of the 4 examples running with the next one of interest being the RandomWriter. more info in this regard is appreciated. lastly, and this isn't a hot topic just yet, but i'm curious as to node discovery capabilities and the prevailing thoughts in this regard, with discovery implying that the namenode server can come up w/ a known set of slaves yet can discover other such candiates at run time. i'm a committer on jxta.org and if there is such interest in exploring discovery ideas i'd love to work with you folks to see if jxta can be one such provider, at least for discovery initially w/ perhaps overall wire transport in time. i'd be really happy w/ discovery in as a simple first step. thx for your time, - james
