Why not gnat Ada in gcc?

[Mike Stump]

[ be sure to read to the end for a 10 Mbps dedicated colo offer ]

> Date: Thu, 12 Oct 2000 16:40:49 -0600 (MDT)
> From: Richard Stallman <rms@gnu.org>
> To: law@cygnus.com
> CC: guerby@acm.org, dewar@gnat.com, gcc@gcc.gnu.org

> But it is unreasonable to ask ACT to put its repository onto a machine
> run by Cygnus, and associated by the public with Cygnus.

I'd be more curious if ACT has a problem with this.

> That is nice, but the problems we should solve by moving the
> repository are not a matter of to network bandwidth.  Increasing
> bandwidth to the repository, while not a bad thing, won't help the
> GNU Project in any of the larger ways.

I think the issue of control can be mitigated by ensuring there is
enough non-redhat people in the decision making process.  I think what
was done with the steering committee is a fine example of how to do it
correctly.  I'd leave the decision about who runs/owns and where the
machine is to be determined by the steering committee with input from
all.  I think we can trust them to make the right decision.  I think
who makes sure it it powered isn't all that important (as long as it
is powered).  The factors I consider important are, who can decide
what is on our web page, who can decide who can have write access into
the cvs repository, who sets policy as to what mailing list software
we run, who decides list policies, who decides what name the web
server will have, who determins what advertising the site can have,
who determines who can mirror, who determines who can have accounts
and so on.  As long as that appears to be fair, I think we are in a
pretty good position.

A few years ago, before having history with the currect site, I
worried about the server.  I wanted to make use of gcc in a critical
way at my new company, and to do this, I needed slightly more
guarantees about the cvs respository than I had.  I solved the problem
by asking for an account on the cvs machine and mirroring the whole
cvs repository.  That was the main thing I had to have a hard
guarantee about, and by mirroring it, I had what I needed.


To throw in addition possibilities into the mix, I can offer a
dedicated 10 Mbps connection that is of `excellent quality'.  The site
has a 622 connection to the Internet currently, so we aren't talking
about an overloaded connection.  Think, the type of quality you would
need to run the busiest root name server on the Internet, in fact,
that is where I am thinking.  The only problem with this, it would be
in RedWood city, and touching the machine, should it be necessary
would be harder than it is now.  If we want more network bandwidth,
that is an excellent way to go.  To see the new digs, run a traceroute
to f.root-servers.net, www.xfree86.org or www.netbsd.org.  The last
two took up a similar offer of connectivity.  We could talk with them
to find out their actual experience and issues, if we wanted to.

A couple of beefy traceroutes from my site:

bash$ t -q 200 -Q www.netbsd.org
traceroute to nbwww.isc.org (204.152.186.171), 30 hops max, 40 byte packets
 1  cr1-hfc4.cstvl1.sfba.home.net (24.1.125.1)  (17.3 ms/21.7 ms(+-1.7 ms)/141.2 ms) 200/200 (0.00%)
 2  r1-ge1-0-1000bt.frmt1.sfba.home.net (24.1.80.1)  (6.1 ms/14.7 ms(+-1.1 ms)/34.4 ms) 200/200 (0.00%)
 3  10.0.255.109 (10.0.255.109)  (6.2 ms/13.6 ms(+-1.0 ms)/31.8 ms) 200/200 (0.00%)
 4  c1-pos6-1.snjsca1.home.net (24.7.74.61)  (7.8 ms/13.8 ms(+-1.0 ms)/32.0 ms) 200/200 (0.00%)
 5  bb2-pos1-0.paix.nap.home.net (24.7.74.170)  (7.2 ms/13.2 ms(+-1.0 ms)/33.4 ms) 200/200 (0.00%)
 6  paix2a.head2.pla.mibh.net (198.32.175.4)  (6.7 ms/15.5 ms(+-1.2 ms)/52.2 ms) 200/200 (0.00%)
 7  pla1b.head4.pla.mibh.net (204.152.184.214)  (8.0 ms/14.1 ms(+-1.0 ms)/42.9 ms) 200/200 (0.00%)
 8  head1.rwc.mibh.net (128.177.255.7)  (8.4 ms/14.8 ms(+-1.1 ms)/38.5 ms) 200/200 (0.00%)
 9  nbwww.isc.org (204.152.186.171)  (7.9 ms/16.6 ms(+-1.3 ms)/61.0 ms) 200/200 (0.00%)

bash$ t -q 200 -Q gcc.gnu.org
traceroute to gcc.gnu.org (205.180.83.71), 30 hops max, 40 byte packets
 1  cr1-hfc4.cstvl1.sfba.home.net (24.1.125.1)  (11.1 ms/24.7 ms(+-1.9 ms)/65.3 ms) 200/200 (0.00%)
 2  r1-ge1-0-1000bt.frmt1.sfba.home.net (24.1.80.1)  (9.0 ms/18.6 ms(+-1.4 ms)/44.3 ms) 200/200 (0.00%)
 3  10.0.255.109 (10.0.255.109)  (7.1 ms/18.8 ms(+-1.5 ms)/63.0 ms) 200/200 (0.00%)
 4  c1-pos6-1.snjsca1.home.net (24.7.74.61)  (7.4 ms/20.8 ms(+-1.6 ms)/61.6 ms) 200/200 (0.00%)
 5  24.7.70.138 (24.7.70.138)  (6.9 ms/15.5 ms(+-1.2 ms)/61.5 ms) 200/200 (0.00%)
 6  p6-0.paloalto-nbr1.bbnplanet.net (4.0.6.97)  (8.5 ms/15.0 ms(+-1.1 ms)/44.0 ms) 200/200 (0.00%)
 7  p1-0.paloalto-cr1.bbnplanet.net (4.0.6.74)  (7.7 ms/14.9 ms(+-1.1 ms)/48.4 ms) 200/200 (0.00%)
 8  p1-0-0.paloalto-cr13.bbnplanet.net (4.0.2.222)  (7.9 ms/14.9 ms(+-1.1 ms)/40.7 ms) 200/200 (0.00%)
 9  s0.cygnus3.bbnplanet.net (4.1.120.86)  (102.8 ms/234.6 ms(+-16.9 ms)/346.8 ms) 200/200 (0.00%)
10  sourceware.cygnus.com (205.180.83.71)  (52.5 ms/220.9 ms(+-18.4 ms)/555.7 ms) 156/200 (22.00%)

The latency for me is 13x better on average, and 9x better worst case,
6x better best case.  During the day, the numbers will be even better.
During a cygwin release, the numbers improve yet some more.