[master] 6c533a053 About our IBM S390x CI client

[Poul-Henning Kamp]

commit 6c533a053c16d6bf76dc9ac0da07a30f487b90f6
Author: Poul-Henning Kamp <phk at FreeBSD.org>
Date:   Mon May 24 14:43:29 2021 +0000

    About our IBM S390x CI client

diff --git a/doc/sphinx/phk/index.rst b/doc/sphinx/phk/index.rst
index cbe4ff314..8897a0662 100644
--- a/doc/sphinx/phk/index.rst
+++ b/doc/sphinx/phk/index.rst
@@ -13,6 +13,7 @@ You may or may not want to know what Poul-Henning thinks.
 .. toctree::
 	:maxdepth: 1
 
+	legacy.rst
 	ip_address.rst
 	vdd19q3.rst
 	quic.rst
diff --git a/doc/sphinx/phk/legacy.rst b/doc/sphinx/phk/legacy.rst
new file mode 100644
index 000000000..914861c05
--- /dev/null
+++ b/doc/sphinx/phk/legacy.rst
@@ -0,0 +1,81 @@
+..
+	Copyright (c) 2021 Varnish Software AS
+	SPDX-License-Identifier: BSD-2-Clause
+	See LICENSE file for full text of license
+
+.. _phk_legacy:
+
+=========
+Legacy-.*
+=========
+
+In the middle of an otherwise pleaseant conversation recently, the
+other person suddenly burst out that "Varnish was part of our
+legacy software."
+That stung a bit.
+
+But fair enough:  They *have* been running varnish since 2009 or so.
+
+Neither Raymond's "New Hacker's Dictionary", nor the legacy publication
+it tried to replace, Kelly-Bootle's "The Devils DP Dictionary", define
+"legacy software".  The former because the collective "we" did not
+bother with utter trivialities such as invoicing, the latter because
+back then people didnt abandon software.
+
+Tomorrow I will be sitting in a small hut in the middle of a field,
+trying to figure out what somebody could possibly have been thinking
+about, when 10 years ago they claimed to implement "V.42" while
+also using their own HDLC frame layout.
+
+"V.42" and "HDLC" are also a legacy at this point, but chances are
+you have used it:  That was the hot way to do error-correction on
+modems, when dialing into a BBS or ISP in the 1990ies.
+
+I guess I should say "legacy-modems" ?  
+
+Big-endianess, storing they bytes the sensible way for hex-dumps, is
+rapidly becoming legacy, as the final old HP and SUN irons are
+finally become eWaste.
+
+Objectively there is no difference of merit between little-endian
+and big-endian, the most successful computers of all time fall
+equally into one or the other, and the consolidation towards
+little-endian is more "It's not important" than anything
+else.
+
+But we still have a bit of code which cares about endianess
+in Varnish, in particular in the imported `zlib` code.
+
+For a while I ran a CI client on a WLAN access point with a
+big-endian MIPS-processor.  But with only 128MB RAM the spurious
+error rate made too much noise.
+
+Nothing has been proclaimed "Legacy" more often and with more force
+than the IBM mainframe, but they are still around, keeping the books
+balanced, like they have for half a century.
+
+And because they were born with variable length data types, IBM
+mainframes are big-endian, and because we in Varnish care about
+portability, you can also run Varnish Cache on your IBM mainframe.
+
+Thanks mainly to Ingvar, there are "s390x" architecture Varnish
+packages if you need them.
+
+So I reached out to IBM's FOSS-outreach people and asked if we could
+borrow a cup of mainframe to run a CI-client, and before I knew it,
+the Varnish Cache Project had access to a virtual s390x machine somewhere
+in IBM's cloud.
+
+For once "In the Cloud" *literally* means "On somebody's mainframe" :-)
+
+I'm not up to date on IBM Mainframe technology, the last one I used
+was a 3090 in 1989, so I have no idea how much performance
+IBM has allocated to us, on what kind of hardware, or what it might
+cost.
+
+But it runs a full CI iteration on Varnish Cache in 3 minutes flat,
+making it one of the fastest CI-clients we have.
+
+Thanks IBM!
+
+Poul-Henning, 2021-05-24