Thread: CRC32 function
Both backend and users may have a nice use of the function.
Nice and fast hashing when one doesn’t need encryption.
Attachment
Ilia Kantor wrote: > Both backend and users may have a nice use of the function. > > > > Nice and fast hashing when one doesn't need encryption. We already have MD5 encryption in the server. Why would someone want CRC32? -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Thu, Aug 25, 2005 at 16:14:04 -0400, Bruce Momjian <pgman@candle.pha.pa.us> wrote: > Ilia Kantor wrote: > > Both backend and users may have a nice use of the function. > > > > > > > > Nice and fast hashing when one doesn't need encryption. > > We already have MD5 encryption in the server. Why would someone want > CRC32? Lower CPU utiliization.
Bruno Wolff III wrote: > On Thu, Aug 25, 2005 at 16:14:04 -0400, > Bruce Momjian <pgman@candle.pha.pa.us> wrote: > > Ilia Kantor wrote: > > > Both backend and users may have a nice use of the function. > > > > > > > > > > > > Nice and fast hashing when one doesn't need encryption. > > > > We already have MD5 encryption in the server. Why would someone want > > CRC32? True, but if there are several million rows, a 32-bit CRC isn't really useful (too many collisions). I can see it useful in a few cases, but not for general usefulness. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
Bruno Wolff III <bruno@wolff.to> writes:
> Bruce Momjian <pgman@candle.pha.pa.us> wrote:
>> We already have MD5 encryption in the server. Why would someone want
>> CRC32?
> Lower CPU utiliization.
Like Bruce, I don't really think there is demand for such a function.
But if we were going to offer it, it at least ought to use the existing
implementation in pg_crc.c, instead of duplicating code yet again.
regards, tom lane
On Thu, Aug 25, 2005 at 18:44:13 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote: > Bruno Wolff III <bruno@wolff.to> writes: > > Bruce Momjian <pgman@candle.pha.pa.us> wrote: > >> We already have MD5 encryption in the server. Why would someone want > >> CRC32? > > > Lower CPU utiliization. > > Like Bruce, I don't really think there is demand for such a function. > But if we were going to offer it, it at least ought to use the existing > implementation in pg_crc.c, instead of duplicating code yet again. Maybe I should have elaborated. I was just responding directly to Bruce's question. I doubt the CPU usage is a big deal in typical use and that that the already available cryptographic hashes have advantages such that I don't expect many people to use CRC32.