Hi, Hagan makes the following remark on his paper "Interpolation methods for yield curve construction":"One approach now advocated in some sources is to interpolate (linearly, say) theinput swap rates to the expiries which are not quoted, and then proceed with acomplete information set. However, this decouples the interpolation procedure fromthe bootstrap procedure, even if the chosen interpolation method here is the same asthe interpolation method that will be used to find rates at points which are not nodesafter the bootstrap is completed. "This remark is regarding the approach of first interpolating all missing swap rates upon which the curve can be constructed by simple bootstrapping.I do agree that this approach has some drawbacks. For example, with this approach, it may be more difficult to ensure arbitrary-free forward rate curves.However, I do not see the practical motivation behind Hagan's remark. At the end, either you take this approach or the usual one which takes a fixed interpolation scheme and calibrate the pillar discount factors to the input market rates, the market rates at the missing pillars are always implied one way or another. If you have certain view on a missing market rate, it is another story. But if not, I do not see what his remark is trying to say. Any idea?Thanks.