Обсуждение: enable-unicode-conversion option?
I think that we have concluded that a bigger executable with the unicode conversion functionality does not have any performance penalty. So I would like to remove --enable-unicode-conversion option so that it is always enabled if --enable-multibye is specified. Any objection? -- Tatsuo Ishii
> I think that we have concluded that a bigger executable with the > unicode conversion functionality does not have any performance > penalty. So I would like to remove --enable-unicode-conversion option > so that it is always enabled if --enable-multibye is specified. > Any objection? Makes sense to me. We already have locale and multibyte. No need for a Unicode one too if we can do it automatically. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026
Tatsuo Ishii wrote: > > I think that we have concluded that a bigger executable with the > unicode conversion functionality does not have any performance > penalty. So I would like to remove --enable-unicode-conversion option > so that it is always enabled if --enable-multibye is specified. > Any objection? Will there be #define's one can set manually to get smaller executable ? I'm contemplating porting pg to PocketPC/WinCE platform and there the size does matter. And as WinCE is pure unicode platform there should be no need for 'conversions'. -------------------- Hannu
Hannu Krosing writes: > I'm contemplating porting pg to PocketPC/WinCE platform and there the > size does matter. And as WinCE is pure unicode platform there should be > no need for 'conversions'. The Unicode support is about 1 MB on disk. If you want to run a server and you don't have the extra 1 MB then you've got problems. Plus, you might have clients connecting that are not pure Unicode. Of course I don't completely understand the setup you have in mind. Is WinCE POSIX-compatible? -- Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter
Peter Eisentraut wrote: > > Hannu Krosing writes: > > > I'm contemplating porting pg to PocketPC/WinCE platform and there the > > size does matter. And as WinCE is pure unicode platform there should be > > no need for 'conversions'. > > The Unicode support is about 1 MB on disk. If you want to run a server > and you don't have the extra 1 MB then you've got problems. A server does not necessaryly mean "A Huge DataWarehousing Server", it may be just a fancy addressbook with some growth potential. > Plus, you might have clients connecting that are not pure Unicode. I may refuse to serve them ;) > Of course I don't completely understand the setup you have in mind. Is > WinCE POSIX-compatible? no. it is an embedded os most compatible to Win32 (an OS found on some Intel x86 based PC's ;) so it would be much easier if a non-cygwin Win32 port was done first. it is run on Compaqs iPAQ and other handheld devices, where memory is still a bit problem (16MB of Flash ROM and 32-64 of RAM in standard configurations, iPAQ has 200MHz StrongARM processor) While that may seem very little now, I've run postgres on much worse hardware only a few years ago (40MHz 486sx with 16Mb ram and 20Mb HDD) The port, if ever done, will require much mucking about in internals and may not be possible inside the main source tree anyway so you should not worry too much if removing --enable-unicode-conversion would make some things much easier. The clients will mostly connect from the same device and quite likely one at a time. As linux is already ported to iPAQ it may be easier to use Linux version of postgres there but even then space consuption is of some concern. OTOH it seems to be a good goal to have Postgres be able to run everywhere where Linux can ;) --------------- Hannu
It seems to be resonable to leave #define UNICODE_CONVERSION somewhere (maybe in pg_config.h). -- Tatsuo Ishii > Peter Eisentraut wrote: > > > > Hannu Krosing writes: > > > > > I'm contemplating porting pg to PocketPC/WinCE platform and there the > > > size does matter. And as WinCE is pure unicode platform there should be > > > no need for 'conversions'. > > > > The Unicode support is about 1 MB on disk. If you want to run a server > > and you don't have the extra 1 MB then you've got problems. > > A server does not necessaryly mean "A Huge DataWarehousing Server", it > may > be just a fancy addressbook with some growth potential. > > > Plus, you might have clients connecting that are not pure Unicode. > > I may refuse to serve them ;) > > > Of course I don't completely understand the setup you have in mind. Is > > WinCE POSIX-compatible? > > no. it is an embedded os most compatible to Win32 (an OS found on some > Intel x86 based PC's ;) so it would be much easier if a non-cygwin Win32 > port was done first. > > it is run on Compaqs iPAQ and other handheld devices, where memory is > still > a bit problem (16MB of Flash ROM and 32-64 of RAM in standard > configurations, > iPAQ has 200MHz StrongARM processor) While that may seem very little > now, I've > run postgres on much worse hardware only a few years ago (40MHz 486sx > with > 16Mb ram and 20Mb HDD) > > The port, if ever done, will require much mucking about in internals and > may > not be possible inside the main source tree anyway so you should not > worry too > much if removing --enable-unicode-conversion would make some things much > easier. > > The clients will mostly connect from the same device and quite likely > one > at a time. > > As linux is already ported to iPAQ it may be easier to use Linux version > of > postgres there but even then space consuption is of some concern. > > OTOH it seems to be a good goal to have Postgres be able to run > everywhere > where Linux can ;) > > --------------- > Hannu >
Tatsuo Ishii wrote: > > It seems to be resonable to leave #define UNICODE_CONVERSION somewhere > (maybe in pg_config.h). That's what I was after. -------------- Hannu