| Message ID | 1464017813-22503-1-git-send-email-abrodkin@synopsys.com |
|---|---|
| State | Accepted |
| Headers | show |
Hello, On Mon, 23 May 2016 18:36:53 +0300, Alexey Brodkin wrote: > It turned out for quite some time CDN-powered mirror > of kernel.org is available. See this announce: > https://www.kernel.org/introducing-fastly-cdn.html > > This change switches to that faster mirror by default. > > Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> > Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > Cc: Peter Korsgaard <peter@korsgaard.com> That's a good change. It was suggested some time ago on IRC, and we should do it. > config BR2_KERNEL_MIRROR > string "Kernel.org mirror" > - default "https://www.kernel.org/pub" > + default "https://cdn.kernel.org/pub" > help > kernel.org is mirrored on a number of servers around the world. > The following allows you to select your preferred mirror. > @@ -236,6 +236,9 @@ config BR2_KERNEL_MIRROR > http://www.XX.kernel.org/pub (XX = country code) > http://mirror.aarnet.edu.au/pub/ftp.kernel.org Should this list of mirrors be removed? > + Now when kernel.org is mirrored by CDN as well we'll be using > + it by default to speed-up downloads all over the world. I'm sorry, but I don't understand this sentence. Thomas
>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > Hello, > On Mon, 23 May 2016 18:36:53 +0300, Alexey Brodkin wrote: >> It turned out for quite some time CDN-powered mirror >> of kernel.org is available. See this announce: >> https://www.kernel.org/introducing-fastly-cdn.html >> >> This change switches to that faster mirror by default. >> >> Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> >> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> >> Cc: Peter Korsgaard <peter@korsgaard.com> > That's a good change. It was suggested some time ago on IRC, and we > should do it. >> config BR2_KERNEL_MIRROR >> string "Kernel.org mirror" >> - default "https://www.kernel.org/pub" >> + default "https://cdn.kernel.org/pub" >> help >> kernel.org is mirrored on a number of servers around the world. >> The following allows you to select your preferred mirror. >> @@ -236,6 +236,9 @@ config BR2_KERNEL_MIRROR >> http://www.XX.kernel.org/pub (XX = country code) >> http://mirror.aarnet.edu.au/pub/ftp.kernel.org > Should this list of mirrors be removed? Does it even make sense to have a configuration option for it if the CDN is used?
Hi Thomas, On Mon, 2016-05-23 at 17:46 +0200, Thomas Petazzoni wrote: > Hello, > > On Mon, 23 May 2016 18:36:53 +0300, Alexey Brodkin wrote: > > > > It turned out for quite some time CDN-powered mirror > > of kernel.org is available. See this announce: > > https://www.kernel.org/introducing-fastly-cdn.html > > > > This change switches to that faster mirror by default. > > > > Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> > > Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > > Cc: Peter Korsgaard <peter@korsgaard.com> > That's a good change. It was suggested some time ago on IRC, and we > should do it. > > > > > config BR2_KERNEL_MIRROR > > string "Kernel.org mirror" > > - default "https://www.kernel.org/pub" > > + default "https://cdn.kernel.org/pub" > > help > > kernel.org is mirrored on a number of servers around the world. > > The following allows you to select your preferred mirror. > > @@ -236,6 +236,9 @@ config BR2_KERNEL_MIRROR > > http://www.XX.kernel.org/pub (XX = country code) > > http://mirror.aarnet.edu.au/pub/ftp.kernel.org > Should this list of mirrors be removed? Well I have no preference here. I mean user still may have a preferred mirror so why would we limit him or her here? > > > > + Now when kernel.org is mirrored by CDN as well we'll be using > > + it by default to speed-up downloads all over the world. > I'm sorry, but I don't understand this sentence. If I'm not mistaken CDN allows to use the best server for each user (based on some criteria). Now kerenl.org is mirrored with use of that technology. And in theory users of Buildroot all over the world will benefit from much faster downloads from this CDN mirror (which will be in real life different physical servers for different users). Not sure if that explanation makes more sense. But for me use of CDN gave 2.5 boost (1.8 vs 5 Mbit/sec). And still there's a probability that it was our ISP that limited to 5 Mbit'sec and given I have a wider channel download speed might be even higher. -Alexey
Hello, On Mon, 23 May 2016 17:57:01 +0200, Peter Korsgaard wrote: > > Should this list of mirrors be removed? > > Does it even make sense to have a configuration option for it if the CDN > is used? I asked the exact same question when a person came on IRC to suggest this change. I referred to what we did for the SourceForge situation. But apparently, this CDN thing is less smart than the SourceForge thing in that it might redirect you to a non-working mirror (i.e it doesn't check). However, the person on IRC was also suggesting to do the change for BR2_GNU_MIRROR, so maybe the case is different. So, I would suggest to change the default values for now, leave it this way for a while, and if nobody complains of failed downloads, we can remove the option completely. Thomas
Hi Thomas, On Mon, 2016-05-23 at 18:04 +0200, Thomas Petazzoni wrote: > Hello, > > On Mon, 23 May 2016 17:57:01 +0200, Peter Korsgaard wrote: > > > > > > Should this list of mirrors be removed? > > > > Does it even make sense to have a configuration option for it if the CDN > > is used? > I asked the exact same question when a person came on IRC to suggest > this change. I referred to what we did for the SourceForge situation. > > But apparently, this CDN thing is less smart than the SourceForge thing > in that it might redirect you to a non-working mirror (i.e it doesn't > check). > > However, the person on IRC was also suggesting to do the change for > BR2_GNU_MIRROR, so maybe the case is different. > > So, I would suggest to change the default values for now, leave it this > way for a while, and if nobody complains of failed downloads, we can > remove the option completely. Agree only real life test may answer how reliable that thing is. But given there's an obvious benefit for users and current BR state we'll have plenty of time to try this out and if people start complaining we'll revert back to a simple www.kernel.org before the next rc1. -Alexey
Hello, On Mon, 23 May 2016 16:04:05 +0000, Alexey Brodkin wrote: > > > http://www.XX.kernel.org/pub (XX = country code) > > > http://mirror.aarnet.edu.au/pub/ftp.kernel.org > > Should this list of mirrors be removed? > > Well I have no preference here. > I mean user still may have a preferred mirror so why would we > limit him or her here? Well, if we're pointing to a CDN then really there is no longer any reason to point users to www.XXX.kernel.org with XX = country code, since that's what a CDN is doing automatically. > > > + Now when kernel.org is mirrored by CDN as well we'll be using > > > + it by default to speed-up downloads all over the world. > > I'm sorry, but I don't understand this sentence. > > If I'm not mistaken CDN allows to use the best server for each user > (based on some criteria). Now kerenl.org is mirrored with use of that > technology. And in theory users of Buildroot all over the world will > benefit from much faster downloads from this CDN mirror (which will > be in real life different physical servers for different users). Yes, of course. But I still think the wording is wrong. The wording feels like something for the commit log, not for the help text. A correct wording for the help text could be: The default value points to kernel.org's CDN, which will automatically redirect you to a mirror geographically close to you. > But for me use of CDN gave 2.5 boost (1.8 vs 5 Mbit/sec). > And still there's a probability that it was our ISP that limited to > 5 Mbit'sec and given I have a wider channel download speed might be > even higher. Right. Thanks, Thomas
Hi Thomas, On Mon, 2016-05-23 at 19:18 +0200, Thomas Petazzoni wrote: > Hello, > > On Mon, 23 May 2016 16:04:05 +0000, Alexey Brodkin wrote: > > > > > > > > > > > > > > http://www.XX.kernel.org/pub (XX = country code) > > > > http://mirror.aarnet.edu.au/pub/ftp.kernel.org > > > Should this list of mirrors be removed? > > Well I have no preference here. > > I mean user still may have a preferred mirror so why would we > > limit him or her here? > Well, if we're pointing to a CDN then really there is no longer any > reason to point users to www.XXX.kernel.org with XX = country code, > since that's what a CDN is doing automatically. Indeed but given our previous discussion it makes sense to leave those help strings for time being. > > > > > > > > > > > > > + Now when kernel.org is mirrored by CDN as well we'll be using > > > > + it by default to speed-up downloads all over the world. > > > I'm sorry, but I don't understand this sentence. > > If I'm not mistaken CDN allows to use the best server for each user > > (based on some criteria). Now kerenl.org is mirrored with use of that > > technology. And in theory users of Buildroot all over the world will > > benefit from much faster downloads from this CDN mirror (which will > > be in real life different physical servers for different users). > Yes, of course. But I still think the wording is wrong. The wording > feels like something for the commit log, not for the help text. A > correct wording for the help text could be: > > The default value points to kernel.org's CDN, which will > automatically redirect you to a mirror geographically close to > you. Do you want me to resend the patch or you'll do this fixup before patch application? -Alexey
>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > Hello, > On Mon, 23 May 2016 17:57:01 +0200, Peter Korsgaard wrote: >> > Should this list of mirrors be removed? >> >> Does it even make sense to have a configuration option for it if the CDN >> is used? > I asked the exact same question when a person came on IRC to suggest > this change. I referred to what we did for the SourceForge situation. > But apparently, this CDN thing is less smart than the SourceForge thing > in that it might redirect you to a non-working mirror (i.e it doesn't > check). :/ > However, the person on IRC was also suggesting to do the change for > BR2_GNU_MIRROR, so maybe the case is different. > So, I would suggest to change the default values for now, leave it this > way for a while, and if nobody complains of failed downloads, we can > remove the option completely. Ok. Missing files should normally fall back to sources.b.o, so we might not hear about it though.
Hello, On Mon, 23 May 2016 18:36:53 +0300, Alexey Brodkin wrote: > It turned out for quite some time CDN-powered mirror > of kernel.org is available. See this announce: > https://www.kernel.org/introducing-fastly-cdn.html > > This change switches to that faster mirror by default. > > Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> > Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> > Cc: Peter Korsgaard <peter@korsgaard.com> > --- > Config.in | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) Applied to next after slightly tweaking the help text, thanks! Thomas
diff --git a/Config.in b/Config.in index 93b4f69..71f0161 100644 --- a/Config.in +++ b/Config.in @@ -225,7 +225,7 @@ config BR2_BACKUP_SITE config BR2_KERNEL_MIRROR string "Kernel.org mirror" - default "https://www.kernel.org/pub" + default "https://cdn.kernel.org/pub" help kernel.org is mirrored on a number of servers around the world. The following allows you to select your preferred mirror. @@ -236,6 +236,9 @@ config BR2_KERNEL_MIRROR http://www.XX.kernel.org/pub (XX = country code) http://mirror.aarnet.edu.au/pub/ftp.kernel.org + Now when kernel.org is mirrored by CDN as well we'll be using + it by default to speed-up downloads all over the world. + config BR2_GNU_MIRROR string "GNU Software mirror" default "http://ftp.gnu.org/pub/gnu"
It turned out for quite some time CDN-powered mirror of kernel.org is available. See this announce: https://www.kernel.org/introducing-fastly-cdn.html This change switches to that faster mirror by default. Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Peter Korsgaard <peter@korsgaard.com> --- Config.in | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)