Stephen, Perhaps it is time to fix iproute2 interpretation of kbit/mbit/gbit. Currently, they are interpreted as powers of 2 (i.e. 10mbit = 10*1024*1024), which is absolutely incorrect when dealing with networking, as line speeds are always interpreted in decimal. Example: 10Mbit ethernet is 10 000 000 bits/second. Someone who may be trying to rate-limit outbound traffic is bound for a surprise when tc''s 10mbit does not match physical line characteristic. Other examples: 28k modem is 28000 bit/s, 56k is 56000, OC-3 SONET (155Mbit) is 155000000 bit/s, etc. There isn''t a technology that is quoted with kbits meaning 1024bit/s. -alex On Tue, 8 Jun 2004, Stephen Hemminger wrote:> A new version of the iproute2 utilities is available to handle the new > extensions for 2.6.7. > * Based on the last known good version of iproute2 from Alexy > * Included most of the vendor patches (except for the stupid ones). > * Got rid of lots of the glibc workarounds, I intend this to build > and run on 2.6 (and 2.4) only. > * Fixed some parsing and formatting bugs. > * Added gigabit as a rate. > * Added HTB and delay scheduler > * Added support for new tcp_info extensions to ss > > The website is: > http://developer.osdl.org/dev/iproute2 > and the download is in: > http://developer.osdl.org/dev/iproute2/download > > This version builds with 2.6.7 as the kernel include files, so either have > the files in /usr/include/linux up to date or modify the top level Makefile > to point to a kernel build. Will workout a way to build on 2.4 next. > > - > To unsubscribe from this list: send the line "unsubscribe linux-net" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >- To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
> But should we break existing scripts?? One possibility would be to make > things case dependant (K = 1024 and k = 1000) or something like that.That does makes sense for entering data, however, for display of results, they need to be in one format. How about a global flags, -k/-K which would change all multipliers (for input as well as output) to 1000 or 1024? That way, nothing existing will be broken, and a warning should be placed in documentation that without -k flag, results are misleading. I''ll leave the default selection for -k/-K up to you. Alternatively, (tc -k <command>) -alex _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
A new version of the iproute2 utilities is available to handle the new extensions for 2.6.7. * Based on the last known good version of iproute2 from Alexy * Included most of the vendor patches (except for the stupid ones). * Got rid of lots of the glibc workarounds, I intend this to build and run on 2.6 (and 2.4) only. * Fixed some parsing and formatting bugs. * Added gigabit as a rate. * Added HTB and delay scheduler * Added support for new tcp_info extensions to ss The website is: http://developer.osdl.org/dev/iproute2 and the download is in: http://developer.osdl.org/dev/iproute2/download This version builds with 2.6.7 as the kernel include files, so either have the files in /usr/include/linux up to date or modify the top level Makefile to point to a kernel build. Will workout a way to build on 2.4 next. - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, 8 Jun 2004 18:32:09 -0400 (EDT) alex@pilosoft.com wrote:> Stephen, > > Perhaps it is time to fix iproute2 interpretation of kbit/mbit/gbit. > > Currently, they are interpreted as powers of 2 (i.e. 10mbit = > 10*1024*1024), which is absolutely incorrect when dealing with networking, > as line speeds are always interpreted in decimal. > > Example: 10Mbit ethernet is 10 000 000 bits/second. Someone who may be > trying to rate-limit outbound traffic is bound for a surprise when tc''s > 10mbit does not match physical line characteristic. > > Other examples: 28k modem is 28000 bit/s, 56k is 56000, OC-3 SONET > (155Mbit) is 155000000 bit/s, etc. There isn''t a technology that is > quoted with kbits meaning 1024bit/s.But should we break existing scripts?? One possibility would be to make things case dependant (K = 1024 and k = 1000) or something like that. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Am Wednesday 09 June 2004 01:31 schrieb Stephen Hemminger:> But should we break existing scripts??Interesting questions - would it break scripts, or fix them? How many people did actually check wether k stands for 1024 or 1000 when writing their scripts?> One possibility would be to make things case dependant (K = 1024 and k > 1000) or something like that.If 1024 is not used in real networks (I don''t know) then you should just ditch it completely, IMO. Offering two possibilities is fine, but could also easily lead to even more confusion. Andreas _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hi. Andreas Klauer wrote:> Interesting questions - would it break scripts, or fix them? How many > people did actually check wether k stands for 1024 or 1000 when writing > their scripts?I did. Bye, Mike _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Tue, Jun 08, 2004 at 04:10:36PM -0700, Stephen Hemminger wrote:> A new version of the iproute2 utilities is available to handle the new > extensions for 2.6.7. > * Added HTB and delay scheduler > * Added support for new tcp_info extensions to ssWhy are TC_CONFIG_DIFFSERV=n and TC_CONFIG_ATM=n in Config? Is there a reason for not building them by default!?? btw, perhaps you could integrate some other TC patches like wrr and esfq.> This version builds with 2.6.7 as the kernel include files, so either have > the files in /usr/include/linux up to date or modify the top level Makefile > to point to a kernel build. Will workout a way to build on 2.4 next.I guess this is the reason that compiling in "misc/" fails with my 2.6.6 headers. -- damjan | дамјан This is my jabber ID --> damjan@bagra.net.mk <-- not my mail address!!! _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Stephen Hemminger wrote:> A new version of the iproute2 utilities is available to handle the new > extensions for 2.6.7.Cool, thanks!> * Based on the last known good version of iproute2 from Alexy > * Included most of the vendor patches (except for the stupid ones).Why not include the redhat patch to allow the ip command to control the promiscuous mode of an interface? cheers, Pádraig. - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, 9 Jun 2004 02:31:29 +0200 Andreas Klauer <Andreas.Klauer@metamorpher.de> wrote:> If 1024 is not used in real networks (I don''t know) then you should > just ditch it completely, IMO. Offering two possibilities is fine, but > could also easily lead to even more confusion.Is not only not used in networks, but also not used in hard drives and other storage media, so when you get a 30GB hard drive you only get 30 000 000 Bytes. The correct way according to SI (Internationa System of Units) and the IEC (International Electrotechnical Commission) are explained here [1] So if you''re using power of two factors (times 1024) they should be called kibi (Ki), mebi (Mi), gibi (Gi), tebi (Ti), pebi (Pi), exbi (Ei), and for power of ten they should be kilo (k), mega (M), giga (G), etc. Notice the lowercase k for kilo, as uppercase K is for Kelvin units in the SI. If well documented and frontend developers are encouraged to use the correct unit symbols I don''t see any major problems in the mid term, only some at first until people gets used to them. The kind of users that will usually use iproute should have the brain to understand that Mbps is 1000 kbps wich is 1000 bits per second and also that MBps is 1000 kBps wich is 1000 bytes per second, and to understand that if they want to use power of two they will have Mibps that is 1024 Kibps wich are 1024 bits per second. [1] http://en.wikipedia.org/wiki/Mebi Federico Figueroa C. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Tuesday, 08 June 2004, at 18:42:46 -0400, alex@pilosoft.com wrote:> That does makes sense for entering data, however, for display of results, > they need to be in one format. >That of units, their use and meaning is a common topic at some places, including linux-kernel. There seems to be several international standards with respect to units, both base-2 (2^n) an base-10 (10^n). Maybe you have around units(7), a man page that gives a bit of information on the subject. In my opinion, data should always be shown correct, that is, 1048576 will be eiher 1024 kibi (1024 Ki), or 1048.576 kilo (1048.576 k), but not anything else. The possible change in suffixes could break existing scripts in a (hopefully) minor way, but changing rate specification in "tc" should be thought for longer, the breakage here can be important. Just my two euro cents ;-) -- Jose Luis Domingo Lopez Linux Registered User #189436 Debian Linux Sid (Linux 2.6.7-rc1) - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Changing such base things is easy. Change the package name, iproute2-ng, with all the new and arguably more correct value terminology. David Jose Luis Domingo Lopez wrote:>[...] > >Maybe you have around units(7), a man page that gives a bit of >information on the subject. In my opinion, data should always be shown >correct, that is, 1048576 will be eiher 1024 kibi (1024 Ki), or >1048.576 kilo (1048.576 k), but not anything else. > >The possible change in suffixes could break existing scripts in a >(hopefully) minor way, but changing rate specification in "tc" should be >thought for longer, the breakage here can be important. > >Just my two euro cents ;-) > > >
somebody has tested this version ? is stable ? _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wed, 9 Jun 2004 16:29:25 +0200 Damjan <gdamjan@mail.net.mk> wrote:> On Tue, Jun 08, 2004 at 04:10:36PM -0700, Stephen Hemminger wrote: > > A new version of the iproute2 utilities is available to handle the new > > extensions for 2.6.7. > > * Added HTB and delay scheduler > > * Added support for new tcp_info extensions to ss > > Why are TC_CONFIG_DIFFSERV=n and TC_CONFIG_ATM=n in Config? Is there a > reason for not building them by default!??That''s how it was when I picked it up.> btw, perhaps you could integrate some other TC patches like wrr and > esfq.If people send them they will get integrated. But only if the mainline kernel supports them. I resisted putting in options for things that only show up in vendor or special kernels.> > This version builds with 2.6.7 as the kernel include files, so either have > > the files in /usr/include/linux up to date or modify the top level Makefile > > to point to a kernel build. Will workout a way to build on 2.4 next. > > I guess this is the reason that compiling in "misc/" fails with my 2.6.6 > headers.Fixed in the next snapshot. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 09 June 2004 01:31, Stephen Hemminger wrote:> But should we break existing scripts?? One possibility would be to make > things case dependant (K = 1024 and k = 1000) or something like that.I vote for yes. What we can do, is use an other name so the difference is clear that it''s not tc, but the new tc. Mhh, why don''t call this newtc? Or tcnew? Or tc2? So people can use tc and not change the scripts, or use tcnew and change all scripts accordingly. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hi, Anyone bothered to check : http://en.wikipedia.org/wiki/Kibi this discussion just is too silly. ;-) regards, pieter frans On 9 Jun 2004 at 21:21, Stef Coene wrote:> On Wednesday 09 June 2004 01:31, Stephen Hemminger wrote: > > But should we break existing scripts?? One possibility would be to make > > things case dependant (K = 1024 and k = 1000) or something like that. > I vote for yes. > What we can do, is use an other name so the difference is clear that it''s not > tc, but the new tc. Mhh, why don''t call this newtc? Or tcnew? Or tc2? > So people can use tc and not change the scripts, or use tcnew and change all > scripts accordingly. > > Stef > > -- > stef.coene@docum.org > "Using Linux as bandwidth manager" > http://www.docum.org/ > _______________________________________________ > LARTC mailing list / LARTC@mailman.ds9a.nl > http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/ >------------------------------000000--------------------------------- , , / \ ((__-^^-, -^^-__)) `- _---'' `---_ -'' `---|o` ''o|---'' \ ` / ): :( :o_o: "-" Support Free Software : http://www.gnu.org/fsf/fsf.html and sign the petition : petition.eurolinux.org/index_html _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wed, 9 Jun 2004 21:21:52 +0200 Stef Coene <stef.coene@docum.org> wrote:> On Wednesday 09 June 2004 01:31, Stephen Hemminger wrote: > > But should we break existing scripts?? One possibility would be to make > > things case dependant (K = 1024 and k = 1000) or something like that. > I vote for yes. > What we can do, is use an other name so the difference is clear that it''s not > tc, but the new tc. Mhh, why don''t call this newtc? Or tcnew? Or tc2? > So people can use tc and not change the scripts, or use tcnew and change all > scripts accordingly.I don''t want to fork or make the iproute2 utilities explicitly different from earlier versions. Do we really want to repeat the mess of proc tools. But, in this case the change is actually small and contained, I haven''t heard one person who would be upset by switching. So the next snapshot will use 1000''s for rates and 1024 for sizes as per the standard usage. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
>I don''t want to fork or make the iproute2 utilities explicitly different from >earlier versions. Do we really want to repeat the mess of proc tools. > >Agree. Be bold. Be unpopular. Be different. But whatever, avoid a mass of incompatible versions with dozens of options.... _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hi,
Anyone bothered to check :
http://en.wikipedia.org/wiki/Kibi
this discussion just is too silly.
;-)
regards,
pieter frans
------------------------------000000---------------------------------
, ,
/ \
((__-^^-, -^^-__))
`- _---'' `---_ -''
`---|o` ''o|---''
\ ` /
): :(
:o_o:
"-"
Support Free Software : http://www.gnu.org/fsf/fsf.html
and sign the petition : petition.eurolinux.org/index_html
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
On Wednesday 09 June 2004 22:37, Stephen Hemminger wrote:> On Wed, 9 Jun 2004 21:21:52 +0200 > > Stef Coene <stef.coene@docum.org> wrote: > > On Wednesday 09 June 2004 01:31, Stephen Hemminger wrote: > > > But should we break existing scripts?? One possibility would be to > > > make things case dependant (K = 1024 and k = 1000) or something like > > > that. > > > > I vote for yes. > > What we can do, is use an other name so the difference is clear that it''s > > not tc, but the new tc. Mhh, why don''t call this newtc? Or tcnew? Or > > tc2? So people can use tc and not change the scripts, or use tcnew and > > change all scripts accordingly. > > I don''t want to fork or make the iproute2 utilities explicitly different > from earlier versions. Do we really want to repeat the mess of proc tools. > > But, in this case the change is actually small and contained, I haven''t > heard one person who would be upset by switching. > > So the next snapshot will use 1000''s for rates and 1024 for sizes as per > the standard usage.Isn''t there a way to create 1 binary that can work like the old and the new version? Depending on the name how the binary is executed, the ritgh version is choosen. So you have tcnew and a link to it called tc. If tc is executed, the old version is used, if tcnew is executed, the new version is used. I know it sounds messy ... Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/