Statified Sampling

I have an N of 9780. I have two strata and am primarily interested in comparing the two. Strata A has 9490 or 97% of the population in it while Strata B has 290 or 3%. It seems I need a sample size of around 2000. If I use the proportionate method I would need 1,000 from each stratum which I can't do. Can I use all 290 cases in strata B and weight them to come up with the 50%/50% distribution? If I do that then I wouldn't be sampling from B but using the total population--is that okay? But when I read the difference between proportionate and disproportionate sampling it seems that I end up either way with having to weight the cases in B so I'm not sure which method to use or which is best.......thank you in advance. Velma
That's part of my question. With one strata at 290 and the other at 9490 what is the proper sampling technique if my interest is in comparing the stratus with one another?