Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 2 May 2000 16:44:24 -0700 (PDT)
From:      Kris Kennaway <kris@FreeBSD.org>
To:        Satoshi Asami <asami@freebsd.org>
Cc:        ports@freebsd.org
Subject:   Re: ports projects
Message-ID:  <Pine.BSF.4.21.0005021629410.44965-100000@freefall.freebsd.org>
In-Reply-To: <200005021212.FAA46737@silvia.hip.berkeley.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2 May 2000, Satoshi Asami wrote:

>  @  Fetching distfiles from the nearest master site (status: none)
> 
>     same goes for the people in the other sides of the ponds.  Any
>     good ideas?  "ping" all the MASTER_SITES and sort them?  I know

This has long been on my wishlist: what I envision is a once-off
"sorting" process which tests bandwidths to all MASTER_SITEs in a
nominated list of ports and maintains a database sorted by increasing
bandwidth to the destination. This could be regenerated at will by a
bsd.port.mk target.

The problem is accurately estimating bandwidth. Pinging is a crude metric,
but many sites (or their upstream firewalls) block pings, and ICMP traffic
may be down-prioritized by intermediate routers. It also only measures
latency, not bandwidth. However, it might be a good first-order
approximation (it's the method Gozilla! and other download optimizers use
on Win32).

There's also the pathchar and pchar ports which do a much more intensive
estimation of bandwidth, but the downside is it seems to take a long time.
I haven't really played with it so it should be possible to make it quick
enough to use on large numbers of hosts (e.g. by only measuring the packet
sizes used in typical FTP transfers, etc).

On the ports which I maintain I try and order the MASTER_SITEs in some
kind of network distance order from the US since thats where most of the
users are (and a lot of other countries route their traffic through the US
anyway), but obviously that also disadvantages some segment of the
userbase.

>  @  Better handling of restrictions (what if depended port is
>     illegal, is interactive, etc.) during package build (status: none)

OpenBSD have taken steps in this direction by defining a set of variables
which specify whether a given action is permitted (mirroring, putting on
CDROM, building package, etc)

OpenBSD have also done a sweep for port license information, which is
something we've neglected.

They're also in the process of fixing ports so packages can be built as
non-root, by making the port install into a local directory and package
there (this is basically the same thing as PREFIX-cleanliness)

Another item on my wishlist is for ports to respect CFLAGS/CXXFLAGS/CC/CXX
for controlling the build. There are far too many ports which don't
respect all of these though, so I don't think a bento warning would be
effective.

>  @  Find a replacement for myself so I can retire (status: none)
> 
>     Any takers? ;)

:-)

Kris

----
In God we Trust -- all others must submit an X.509 certificate.
    -- Charles Forsythe <forsythe@alum.mit.edu>



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0005021629410.44965-100000>