My two cents: Microsoft has a lot of conversions that happen throughout the year from Oracle to SQLServer. The transition for Oracle folks to get to SQLServer is made easier by having a sequence object.

However, in my opinion, this is a travesty. Separate Sequence Objects have long been the bane of existence for high speed, fully parallel bulk & batch loading paradigms. To introduce this to SQLServer raises serious concerns in these areas.

It also makes it easier to "blow the lid" off of big-int based sequences especially if a single sequence is used in large or huge tables. I hope Microsoft thought about mitigation strategies for all of these things, including clusters, and parallel servers, and bulk loaders. If they didn't think about this, then those projects that take on the use of Sequence Objects rather than identity columns will struggle to adapt as their volumes grow, and their parallelism needs increase.

I've long thought that Oracle should be the one to change, to add identity columns, rather than the other way around. If you look at other major RDBMS players, they pretty much all have identity columns these days: MySQL, Teradata, DB2 UDB, and so on.

But this is just my opinion, however the article is good coverage on a Sequence Object.

Thank-you,

Dan Linstedt