On 1 July 2010 22:08, John F. Eldredge <j...@jfeldredge.com> wrote: > In fact, the technique of having the user select from a list of words, but > actually storing the value as an arbitrary ID (generally numeric), is the > recommended technique in database design. It is called "normalizing the > database". Having the linking value be an ID value means that, should you > want to change the verbal description of the value, for example from soccer > to association_football, you only have to change the value once, in the > lookup table, rather than changing it in thousands of places. >
Actually, normalising a database actually refers to something a little different than this, and as soon as you get out of theory into real world database design you find that full normalisation is not always the best way to do something - you can over-normalise a database. You need to make choices depending on your platform, and how it's used. So lets look at the effects if we made this change, and decide if we'd like the effects. - We can easily change the description of tags, without having to change the main database. Good, for the reasons above. However, this also means that we can easily corrupt tags - change their meaning from what it was when the tag was used. This would be much easier to do than now, where you have to go do bulk updates. Switch a couple of descriptions around, you could have people entering creeks as paths, brothels as churches (or vice versa), or whatever other mischief you had in mind. And it wouldn't have to be malicious, it could be totally accidental. - Free tag editing would become more dangerous (maybe extinct?) If somebody tags a way as highway=primery, we can figure what they (probably) meant. But if they tag it as 17364 instead of 17634, we have no idea what they actually meant. And if both numbers are in the list, it may be a very hard error to spot. Wether the removal of free tag entry is a good or bad thing would depend on who you ask. - There would be a two step process to using a new tag - create the tag first, then use it. The editors could streamline this, but making it obvious to the user they are creating a new tag may be a good idea. - Having a table of used tags would allow us to add extra metadata to the tags in one central place. We could have translations and localisations, multiple categorisations for a single tag (eg tag X could be under natural, leisure, and landcover), and other options. - There is currently no differentiation between keys that expect free text (name=, note= etc) and keys that expect one of a limited number of options (highway=) We'd still need the free text option for some tags, but would we need to cut it off entirely for others? Stephen _______________________________________________ Tagging mailing list Tagging@openstreetmap.org http://lists.openstreetmap.org/listinfo/tagging
