Re: HTTP/2 Header Encoding Status Update

On Wed, Feb 27, 2013 at 03:12:25PM -0800, James M Snell wrote:
> Using the original epoch, I think we use six bytes currently. The new
> epoch uses five.

Nope. Using an epoch of 1970-01-01T00:00:00Z currently gives a 5-byte
uvarint encoding.

We won't hit 6 bytes until 3058-10-26T03:46:08Z.

I hope we can afford the extra byte by then... :-)

For the record:
* We hit 4 bytes at only ~25 days after whatever epoch we set
* If we decided to make dates be signed varints, we still wouldn't
  hit 6 bytes until 2514.
* If we decide to count milliseconds since the epoch, we're already at 6
  bytes, but that would happen a little over a year after whatever epoch
  we set. (And we wouldn't hit 7 bytes until 2248. Signed milliseconds
  hits 7 bytes in 2109.)

Let's stick to the POSIX epoch.

-- 
Scott Schmit