home *** CD-ROM | disk | FTP | other *** search
- Path: news.ultranet.com!usenet
- From: "Albert P. Belle Isle" <belleisl@cerberus-sys.com>
- Newsgroups: comp.dcom.modems,alt.winsock.trumpet
- Subject: Re: Shouldn't TCPMSS=(MTU-40). ??
- Date: Thu, 29 Feb 1996 10:18:24 -0500
- Organization: Cerberus Systems, Inc.
- Message-ID: <3135C3C0.2E2@cerberus-sys.com>
- References: <4h2kiv$pn5@news2.cais.com> <4h3cvi$sn6@news2.cais.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 2.0 (Win16; I)
-
- jk93 wrote:
- >
- > I'd like to add a "p.s." to this message (an hour later):
- >
- > I know there are many different settings that "will work". I'm mainly
- > wondering if the ISP is just being "sloppy" in giving these settings, or do
- > they know something that i don't know (such as the difference of 40 not
- > being important)?
- >
- > In article <4h2kiv$pn5@news2.cais.com>, jk93@mail.erols.com says...
- > >
- > >My ISP (who has about 20 tech support reps who know very little) tells
- > everyone they should
- > >use the following trumpet settings:
- > >
- > >MTU 576
- > >RWIN 2048
- > >TCPMSS 512
- > >
- > >None of the support reps there knew why the tcpmss is not 40 less than the
- > MTU,
- > >in their recommended setup. But everything i've read elsewhere says it
- > should
- > >be 40 less (I assume it meant exactly 40 less).
- > >Anyone have any idea if those setting are valid?
- > >
-
- jk93:
-
- If you use an MTU the same size as the routers on your access provider's
- Local Area Network, you can set your MSS for TCP transfers (like file or web
- page down-loads) that come from outside his LAN over the Internet to
- anything that will fit in his maximum IP routing packet - i.e. MSS
- *no_larger* than his MTU minus 40 bytes, to leave room for TCP and IP
- headers.
-
- The 40-byte number is the typical header overhead for a single packet. Van
- Jacobson header compression takes advantage of the fact that only some of
- the header info changes from packet-to-packet in a stream of TCP transfers
- to the same recipient (like a file or web-page download). By only sending
- the changes, it reduces the average effective header overhead on such a
- stream from 40 bytes/packet to less than 10. (It's still 40 for single
- control packets.)
-
- The objective is to amortize those header bytes of overhead over as much TCP
- data as possible, without getting router fragmentation of your packets. The
- larger MSS is, the smaller the percentage the header bytes represent.
- However, MSS+40 must fit within the MTU of EVERY router on the IP path
- between the download server and you - not just your ISP's - or they'll be
- fragmented to fit. MSS=536 is 40 bytes less than the Internet default MTU of
- 576, which all IP routers are supposed to accommodate.
-
- This *safe* setting will cause most remote servers to respond with either
- 536- or 512-byte TCP data segments (either of which falls within your
- announced *maximum* segment size of 536 bytes). If you specify a maximum of
- 512, they will always respond with 512, for a slight performance loss.
- Obviously, both sizes will fit in a 576-byte IP transmission unit.
-
- Hope this helps.
-
- Regards,
-
- Al
-
-
- --
- ==================================================================
- Albert P. Belle Isle
- Cerberus Systems, Inc.
-
- Al's Winsock Tuning FAQ -
- http://www.cerberus-sys.com/~belleisl/mtu_mss_rwin.html
- ==================================================================
-