If you get back to what the CDO models are about, the first input is ratings. What we have done and I think I described this in the prior call, is to effectively hair cut ratings, where they have not been moved, so in a previous incarnation, if the ratings have not been cut, we downgraded rating by four notches or six notches for Triple B. We have progressively ramped up the downgrading, where the cups have not been made. In particular, we focused on the mezzanine buckets, because clearly, that is where we discussed at length. There is a fair amount with stress, so what we did was to model out few deals, which did not have mezzanine deals in them and look at the rating degradation of mezzanine classes of those fields. And, if you do that, on realistic assumptions, I stress assumptions, declarations, virtues and so forth. You can come up with good practices of what the appropriate downgrade might be, in the absence of operating agency move. For example, our basic metric is that if you have a, single A rated piece of mezzanine collateral in the CDO bucket that has not yet being re-rated. We would incorporate that into the analysis at a CAA2-type level. We really, substantially have the projected downgrades if you will, within the analysis. We have also spent a fair amount of time on the correlation and on severity numbers. You may recall, at underwriting, I think we used a correlation at 90%. We have opted that significantly and we have chosen to take quite a lot of care into how we differentiate between deals. So, what we have actually done is to go into all the high great deals, split out the vintage distribution of those deals, and use the vintage distribution as a guide to the extent to which we should opt the correlation, assumptions, and also increase the severity assumptions. So, for example, where we believe we got a highly correlated goal, the vintage dispersion is second half of ’05 and almost predominantly, we have certainly doubled sometimes more, and the type of correlation numbers that we have input and equally, we have cut the severity numbers. So, what we tried to do in summary, is to look at what is happening in the market and be thoughtful especially about the mezzanine buckets, and distinguish between transactions in respect of certain key assumptions using vintage distribution as a key moniker in taking those decisions.