[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: OT: Re: 2 Questions



On Thu, 19 Aug 2004 18:35:08 -0400, you wrote:

>On Thu, Aug 19, 2004 at 02:25:08PM -0700, rickie kerndt wrote:
>> At 11:27 AM 8/19/2004, you wrote:
>> >Because MS Office running under wine is way better than openoffice,
>> >which is unbearably slow even on a 2GHz machine, nevermind something a
>> >little older.  Gnumeric and abiword are better, but aren't always able
>> >to make sense of MS Office documents.
>> 
>> Intereting as I use OpenOffice (under FreeBSD w/Gnome desktop) on a 1.7GH 
>> AMD desktop and a 1.4GHz Pentium M laptop. Don't have any issues with speed.
>> 
>
>And?  People say the same thing about mozilla too, and it doesn't mean
>anything.  Just because you are more patient than other people, doesn't
>mean they are wrong for seeking a faster alternative.  MS Office is
>WAY faster than open office, that is the major reason I have seen people
>using MS office under wine instead of OO.
>
>Adam

Adam,

Your points on speed of MS applications, in particular start times and
general responsiveness, are valid but do you understand how MS is doing
this or what costs are involved?

I suggest you read up on PGO (Profile Guided Optimization) and do a ton
of work disassembling some of these supposedly "fast" applications
before posturing that their illusion of "speed" is actually a good
thing. Yes, the form of optimization used does shorten access times and
code paths for the "most used" features but it comes at a significant
price.

As soon as a user does something with the application that is not
"standard" (i.e. something that was sacrificed to enable the supposed
"speed" on more common usage), the illusion of speed disappears. In
other words, those who choose to use an application in particular but
common ways are rewarded with a "boost" in speed. Unfortunately, other
users who need to work in non-standard ways are punished with a
"slow-down" for not following the rest of the herd.

Who gets to decide standard usage, or better said rewards and
punishments, for the operation of open source programs?

Though I've seen no work on the security issues caused by PGO, the
method it uses, namely cutting assembly sub-routines into chunks and
then rearranging the chunks based on usage (including mixing chunks from
various sub-routines), makes debugging and security auditing of binaries
a major pain in the ass.

For the end user who cares nothing about security or reliability, PGO
and other "speed" tricks are a Good Thing (tm) but it requires owning
the source to the OS to keep the tricks keep working. It also requires
significantly more development, testing and debugging time. If you've
got money to burn (as Microsoft does), keeping the illusion of speed for
your products is worth the costs.

Can you name any open source operating system or application so well
funded it has time/money to burn?

Speed tricks are all fun and well for end users but they come at a
serious cost to those who write, test, debug and maintain the code. If
you think the illusion of speed on common usage is actually worth the
effort and cost, they you should start working on it or paying someone
to work on it rather than just complaining about the greater patience
level of those who actually understand both the whole situation and the
illusion.

JCR



Visit your host, monkey.org