Classification doom and gloom

The more I look at this objective the more I rue the day I wrote it. Initially, I thought that it would just involve splitting up some of the larger groups into sub-groups to facilitate browsing but the more I look at the classification (and I think more importantly the more I understand about the subject) the more I realise that there are some fundamental errors with the system. The first is the choice of system - it's based on a "book data" classification which must have been published in the late 80's/ early 90's. It hasn't worn so well in my opinion as computing has moved in ways that could have never been predicted back then. As subjects have come and gone the classification has become fragmented in some areas (for example web development is now the best part of a bay away from web design).
I have the benefit of hindsight but it occurs to me that this classification never had the legs for the long haul. It names specific products for example. Some of which have long gone and are now taking up valuable space in the classification even though there are no books of the shelf. Maybe it's my science background coming back to haunt me but I think I'd prefer it if the classification moved from general to specific in a more graduated way - for example a number for Computing - Applications - Databases and then use the decimal part for specific products. Same for Computing - Programming - Programming Language Type (e.g. Scripting Languages) - Specific Language. This poses a number of questions:

1) Can it be done?
2) How can it be done with the least negative impact on users?
3) Should it be done? Would time be better spent on other things?

The first two are fairly easy to answer. Yes, of course it can be done. The current numbering system could be adapted fairly easily to move in a better way from general to specific so that similar topics are grouped together and are not fragmented. It's also possible to have groups specific enough so that they are not too big to browse using the current numbering BUT re-classifying on the scale that's needed will mean significant moving of stock. How to minimise the impact? I think it would be essential to do as much of the work without moving stock as possible - re-writing the classification, mapping stock to new numbers, and producing the new numbers can all be done without touching the stock. What would be needed then is to pick a really quiet time to reclassify the library - it would have to be Christmas. Then it would be a case of working through the catalogue shelf by shelf and re-stickering the books, editing the catalogue and putting messages on stock that's out to change the stickers on return.

Should it be done? Don't know. It's clear that some reclassification is needed but can I justify a complete reclassification?