home *** CD-ROM | disk | FTP | other *** search
- Path: news.ultranet.com!usenet
- From: "Albert P. Belle Isle" <belleisl@cerberus-sys.com>
- Newsgroups: alt.winsock,alt.winsock.trumpet,comp.dcom.modems
- Subject: Re: 16 bit vs 32 bit winsock
- Date: Fri, 05 Apr 1996 09:21:55 -0500
- Organization: Cerberus Systems, Inc.
- Message-ID: <31652C83.54ED@cerberus-sys.com>
- References: <3157fac9.75178070@146.125.4.24> <315bcd51.5996084@news.iprolink.ch> <4jsbq1$7iu@earth.superlink.net> <31627BA5.BB4@cerberus-sys.com> <4jupph$2pq@boris.eden.com>
- NNTP-Posting-Host: apb-p5-90.cerberus-sys.com
- Mime-Version: 1.0
- Content-Type: text/plain; charset=us-ascii
- Content-Transfer-Encoding: 7bit
- X-Mailer: Mozilla 2.01 (Win16; I)
-
- Alun Jones wrote:
- >
- > In article <31627BA5.BB4@cerberus-sys.com>
- > Albert P. Belle Isle <belleisl@cerberus-sys.com> wrote:
- > > At the risk of getting caught in a WinSock religious flame war, I think the
- > other
- > > readers in the group should be reminded that both the Win95 and Trumpet TCP/IP stacks,
- > > indeed *all* dial-up WinSocks, are *one* bit software on the end that determines how
- > > fast you can surf the 'net: your com port.
- >
- > However, the other end is important, too. 16-bit applications (e.g. Trumpet
- > Winsock), using 16-bit calls to the Windows layer, take up a good portion of
- > the system's time that could perhaps better be served by a 32-bit application,
- > which may be pre-emptively multi-tasked. Thus, users of Win95 will often find
- > that performance improves overall when using the Win95 winsock stack, and 32-
- > bit applications, than using the Trumpet winsock stack (even with 32-bit
- > applications).
- >
-
- Alun:
-
- The above-quoted posting was in response to postings reporting numerical results
- demonstrating faster download performance from Trumpet WinSock than from the Win95
- stack. I was merely offering a reminder that the difference in overhead (with or
- without protected mode context-switching) of updating multiple windows while waiting
- for data through a serial link had little possibility of making the modem faster in
- obtaining that data <g>.
-
- > Many users do other things when they're connected to the internet. What many
- > people fail to realise in the whole 32-bit/16-bit debate
- > is not that the 16-bit apps are slower in themselves, but that they tend to
- > produce a reduction in performance of all the other programs going on around
- > them. In other words, if you only run 16-bit applications, and you don't care
- > about other programs running on your machine, you may be perfectly happy, but
- > if you run other programs, those other programs may slow down. If you hop
- > into Word while an FTP transfer is going on in the background, you don't want
- > that transfer deciding unilaterally how fast Word responds to you.
- >
-
- Well-stated. What's more, once one can eliminate all 16-bit programs, the exclusive use
- of 32-bit applications avoids the context-switching overhead that exacerbates the
- possibility of com overruns. Of course the lack of notification by the Win95 winsock
- that overruns are occuring requires the use of log files to track them down, in order
- to ensure a clean PPP or SLIP layer.
-
- In addition, the Win95 stack's inability to throttle TCP MSS in order to avoid
- fragmentation slowdown except by squeezing IP MTU has made winsock tuning at the TCP/IP
- layers more difficult for some of the posters. Their resulting losses of transfer speed
- below even the low limit imposed by a modem link have led some people to stick with the
- 16-bit stacks for the time being.
-
- Hopefully, the upgrade from Win4.0/DOS7.0 (Win95) to Win4.1/DOS7.1 (Win97?) will
- address some of these concerns in MSFT's usual pattern of continuous improvement.
- Perhaps they'll even be able to offer more than the Win4.0 "almost-preemptive
- multitasking on VxD-blocking," but I expect that we're more likely to see the Win95
- GUI-enhanced version of NT (dumping VxDs for kernal-mode drivers and true
- multi-threaded operation) as the long-term "solution" at a correspondingly "enhanced"
- price ;-)
-
- > Another item to note here is that neither the Win95 stack, nor Trumpet, are
- > strictly dial-up winsock stacks - they can be used over local network systems.
- >
- > Alun.
- > ~~~~
-
- Absolutely well-taken point. The ability to completely eliminate modems in favor of
- direct connections offers the greatest possible speed enhancement, for those fortunate
- enough to be able to afford it.
-
- Regards,
-
- Al
-
- --
- ==================================================================
- Albert P. Belle Isle
- Cerberus Systems, Inc.
-
- Al's Winsock Tuning FAQ -
- http://www.cerberus-sys.com/~belleisl/mtu_mss_rwin.html
- ==================================================================
-