Rationale. GRASS GIS is widely used in scientific, private and government sector. Scientific theories, environmental decisions and actions depend on the outcome of spatial analysis performed with GRASS GIS. Any errors in analytical modules might lead to erroneous conclusions and actions based on them. Currently GRASS GIS project has no clearly outlined procedure how to warn GRASS GIS users of found issues in analytical modules that could lead to good looking but erroneous analysis results. Bug reporting system (Trac) contains different kind of issues and is hard to follow for non-developers; other issues are fixed without opening a bug report in Trac.
Proposal. GRASS GIS could publish an errata document listing found issues in official releases. PSC would make a final decision if issue is serious enough to announce an erratum. Only issues causing meaningful but incorrect results should get an erratum name/number and announcement. An RFC should be adopted to clearly outline procedure and criteria of issuing a GRASS GIS erratum. Following steps should send out a clear message that GRASS project is taking quality issues seriously and should be considered a choice #1 for high quality spatial analysis (aka world domination). Target audience would be scientists basing their research on GRASS GIS, GIS analysts. Final remarks. As final decision is made by PSC, it will increase PSC work load. There must be nominated a responsible person for GRASS GIS erratum text preparation and publishing. Probably it could be a nice idea to set up low traffic mailing list for receiving GRASS GIS errata issues. Should be discussed if existing announcement mailing list should be reused. An option of adding a GRASS GIS version entry to ALL GRASS map metadata should be discussed. It would allow to know exact version used to generate or making of last edits of particular map. Will wait for discussion about proposal in general, corrections for RFC draft (my English is not a top one :( ), other potential changes to be done to increase GRASS GIS trustworthiness and quality. GRASS GIS erratum RFC draft Any bug reporter or developer can nominate a bug for escalating to GRASS GIS erratum issue. Nomination is done by adding a notice to bug at bug tracker or discussing directly at developer mailing list. Any nomination is discussed in developer mailing list to gather necessary information on it's scope, impact, causes and solutions. GRASS PSC evaluates nomination based on information provided by bug report, discussion in developer mailing list and/or other sources. PSC makes a final decision if nominated bug matches criteria of issuing GRASS GIS erratum. Appointed person (WHO?) prepares errata text publishes it. Criteria for bug escalating into erratum. Bug must be present in a official GRASS GIS release. Bug must cause generation of incorrect analysis results that are not so easy to notice. Module crashes or bugs causing easy to identify incorrect results should not be given an erratum. Examples of possible erratum worth bugs are single cell shift of raster result, not enough randomness of expected random module output, loss of output precision due to incorrect floating point handling etc. GRASS GIS erratum text should contain: it's number; date of issue; name(s) of affected module(s); information about affected release(s); a short description of problem; steps resulting in incorrect output (i.e. specific input parameter combination); current state of problem; references to bug report (Trac bug number), developer mailing list thread; any other information relevant to erratum. GRASS GIS errata might receive updates, if it's found to be necessary (i.e. notice of fixing issue, issue scope update etc.). GRASS GIS errata is added to file ERRATUM in all affected branches and any upcoming release branch (trunk). It is published on GRASS GIS web site (Where?). Erratum text and any of it's updates are sent out to ?? mailing list. If erratum has related entry in the bug database, a notice of issuing erratum is added to the bug entry. WBR etc. Maris. _______________________________________________ grass-dev mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/grass-dev
