Phillip Susi wrote: > On 1/23/2012 1:39 PM, H. Peter Anvin wrote: >> GPT enforces its limit because *that's how many slots there are in the >> partition table* (by the way, you *do* handle GPT partition tables with >> more than 128 slots, right?) > > Yes, parted handles more ( or less ) than 128 slots, but only will > create a table with 128 slots. > > The reason I mentioned the gpt limit is because it actually enforces the > limit. Code like that in linux.c assumes that the limit that the > partition table reports is correct, so dos should behave properly and > enforce whatever limit it claims.
The point is that the GPT format dictates we specify a PTE array size in the header, while with a dos partition table, there is no inherent limitation: it's an arbitrarily-long linked list of partition entries. Any PTE-count limitation we impose when writing a dos partition table is artificial or system-specific, and hence undesirable. What if we want to create a partition table that will never be used on a linux-based system? Why should it suffer the limitations of the host kernel?
