I am converting a local table based application to sql server. I am trying to preserve, where possible, my 'legacy' code which relies on seeks, indexes, etc. I am attempting to have essentially the same legacy code operate against remote views, instead of directly against local tables. I have created remote views with the same names as my local tables, except with a v_ prepended. The application needs to create indexes on the remote views 'on the fly'. Originally, I had the buffering mode set to none, 1 or 3?? and the indexes were created without a problem. Then, in some circumstances I began to have 'update conflicts'. As per advice on other threads, I changed the buffering to 5 -now I get the infamous error:
Command cannot be issued on a table with cursors in table buffering mode (Error 1579). I am considering turning buffering off when I create the index then back on again - but I am wary of this - am inclined to believe there might be problems - based upon other threads.

What technique can I use to continue to be able to present my view data with multiple sort sequences - and still use buffering? Maybe I should turn off buffering entirely do my data commits via sql server commit and rollback, instead of tableupdate and tablerevert?

Maybe I can use non-updatable, non-buffered views to present the data the way I want it, and use separate views behind the scenes that are updatable, but without indexes, to actually update the data. But this seems like a lot of work.

Any suggestions would be appreciated.

TIA