[2083] in I/T Delivery
win.mit.edu Developers, November, December
daemon@ATHENA.MIT.EDU (Thomas L. Thornton)
Mon Jan 5 12:54:33 2004
Date: Mon, 5 Jan 2004 12:54:12 -0500 (EST)
Message-Id: <200401051754.i05HsCAP025365@the-rim.mit.edu>
From: "Thomas L. Thornton" <tomt@MIT.EDU>
To: pismere@MIT.EDU, delivery@MIT.EDU, kakapo@MIT.EDU
Cc: phils@MIT.EDU, lerman@MIT.EDU, jud@MIT.EDU, jrw@MIT.EDU, dbaron@MIT.EDU,
azary@MIT.EDU, rar@MIT.EDU, vkumar@MIT.EDU, cporter@MIT.EDU,
ajc@MIT.EDU, patpaul@MIT.EDU, cana@MIT.EDU
Reply-To: tomt@MIT.EDU
Happy 04! Windows developers met fewer times than normal in November
and December, given our extensive reorganization work and holidays, so
this report covers work in both months. It also serves as the
win.mit.edu status report for these periods.
You may be aware of transformations underway in Information Services
and Technology, IS&T, formerly known as Information Systems. These
include reforming Windows development under the new sub-IS&T
organization named Operations and Infrastructure Services, OIS.
Therefore in this process win.mit.edu efforts become a collection of
Windows services and no longer a project. Staff plan to continue
meeting, but aiming to maintain win.mit.edu services. The project
structure of plans, deliverables and reporting wraps up in about a
month. Expect at least another status report before the end of the
project.
Team meeting 5 Nov, 2003, 2:00, W92-125
--------------------------------------
Attendees for this meeting are Richard, Dave, Qing, Joe, Paul and Tom.
Starting with pending business, we review the October Container
Administrators' meeting. See:
http://web.mit.edu/pismere/ctradm/minutes/031031.txt
Old issues arise before the meeting begins - stricter access to event
logs, GP difficulties reported by Stanford's Dennis Michael, an
irreproducible incremental error and Qing's bulk-loading other
domains. We now expect that outgoing.mit.edu will support SMTP auth,
but not require it, in a few weeks. Barry sees KLPR resetting a port,
and we would love to hear more on this to try to reproduce it.
Dave and Paul report on their trips to PDC, mostly Longhorn news.
Indigo is a common communications infrastructure built around Windows
web services, a subsystem for peer-to-peer development. MSBuild.exe
is an XML-based engine, spun out of Apache ANT, that may replace our
wash daily build. The new WinFS storage subsystem has an XML schema
and a relational database view of data. It is NTFS-compatible and has
a transaction-based class layer named WinFX. Paul wonders if this new
FS can be tested against Samba and AFS. Longhorn has a whole new way
of doing identity management for P2P-federated organizations, from the
user's perspective. This *might* enable identity theft, so they want
to provide mitigation through transaction logs. In shorthand, every
client machine will look like a CA and a KDC.
In roundtable Joe gives us a day warning that he will set the
null-session values. He has the next MSDN shipment to index. He
really rocks keeping the production web forms running, especially the
Container Request page.
Qing has a complete source tree building using VS.Net. Like the VS 6
builds, there are two runs, normal and debug. She will examine disk
usage, newer SDKs, scripting the jobs sequentially and setting their
ordering.
Qing has a draft web form protyping a service to track software
licenses. Kakapo should have some results on an inventory and
licensing Discovery project, so they should give us a demo or report.
Now that he is back, Dave will get together with Asanka to plan
torture tests for AFS Unicode and testing AFS under tether and VPN.
Richard examines Terminal Services licensing on W2K and 2K3. He sends
documentation on this to Container Admins and works on a new Joining
WIN document, for which Paul suggests a new mailing list, win-join.
Richard also reviews documentation on the addadmin application, with
some new attention to restricted groups. He develops schedules to
upgrade production servers to SP4 during IAP.
Paul intends to remind PSS about our LSA DC bug report. He plans a
team offsite get-together.
Team meeting 12 Nov, 2003, 2:00, W92-125
---------------------------------------
Informal conversations, no minutes.
Team meeting 19 Nov, 2003, 2:00, W92-125
---------------------------------------
Attendees for this meeting are Dave, Joe, Paul, Qing and Tom.
Tom reminds Joe, Paul and Richard of pending tasks to list MSIs,
implement a transform page and revise the repository spec. We discuss
how to get SIPB upgrading their OSs - Tom will suggest they join a
test public container. With the Drop Date tomorrow, the team must
minimize all domain revisions until IAP. After a group discussion on
process and team plans, we move along to roundtable.
Dave changes incremental to fix ttsp complaints on the syslogs, a
service ticket lifetime issue. Qing sees a related DC patch or manual
configuration setting that she and Richard will examine and resolve.
AFS submounts need more design work and help with debugging from Bob
Basch. CIFS torture testing of IRIPs is under discussion with Asanka.
Qing has finished the license web form, but Richard and others have
precious little time to test it. She tested the DCM on isnt and didnt
given the new ttsp. Paul asks her to look at Chris Cary's OpenAFS MSI
build scripts, to see how to get them into untagged builds. She will
work out with Asanka how to convert his Perl scripts.
Joe has a new version of Blat, v2.3beta. The owners at blat.net have
it and may use it. After finals it will go into distribution, not
using authentication. In related work, he ensures the server popwin
now uses a valid "from" address.
Joe and the team deployed the MS03-049 patch 12 November. About nine
machines have not installed it; they seem unbooted or are always
logged on. A couple XP machines took it multiple times, implying they
need a new XML from MS.
Tom Coveney had an ISO server on which autohotfixer did repetitive 039
patches. Joe talked him through repairs and made sure he now has an
autohotixer workaround.
We plan a GP fix on cross-forest behavior tomorrow. There is a new
remote event log and mirror-distrib on the containers named pismere-*.
The printer installation script now accesses a different key at
bootup. Joe will look at network printer scripts, watching out that
the spooler still may cause hanging.
We have one more Container Admin meeting this year at the end of this
week. See the November/December Container Admin meeting notes:
http://web.mit.edu/pismere/ctradm/minutes/031121.txt
Team meeting 25 Nov, 2003, 2:00, W92-125
---------------------------------------
No meeting.
Team meeting 3 Dec, 2003, 2:00, W92-125
--------------------------------------
Attendees for this meeting are Joe, Paul, Richard, Qing and Tom.
For old business, Paul and Tom ask Kakapo for a review of
recently-completed licensing studies. OpenAFS staff have met about
merging code into OpenAFS.org and are moving forward.
Under new business, Tom notes he has shared WinZephyr source with two
institutions - CERN and University of Maryland. Given the holiday
closing announcement, the team will provide emergency contacts and
schedules, and make sure Container Admins know our plans.
Dave works with Bob Basch on OpenAFS User Interface to some new
backend processes. Asanka has a new document for RX MTU registry keys
and the NetBIOS limit. Qing will put this into the AFS UI. The
default setting of this will drive the CIFS torture test.
Dave has some LPng requests and wonders if he should build a source
distribution, or even how to distribute as a beta binary on campus?
He will consult with the SWRT.
Qing has OpenAFS building as an XML-based installer. She simplifies
and rewrites scripts to put this into the wash head. The installers
work but are 2MB smaller than the Wise-based ones, so she will compare
them carefully. Qing would like to complete a per-container login
report for Theresa, but needs more specification. There are winlogger
errors possibly affecting our Domain statistics tables, and the
database there could be reduced in size - a complete pass through the
DB takes nine hours.
Qing works on a VS automation example which is well underway. She
needs to budget her time and get ready for her early February leave.
Documenting AD data flows will take some of her time, but mostly
Dave's.
Richard tests VPN. He finds that the GP to "Force Local Profiles"
breaks the WIN logon-before script, and will talk to Nathan about it.
Paul describes the External Review Committee report as it stands since
he, Phil and Tom worked out a final copy to share with Jerry and
Vijay. We expect some short version of the report to be published but
are not sure when. Paul plans a CSG trip early in 2004. He will look
into Tech (Corporate) Time problems and contacting associates in IETF.
Joe describes the new machine name stamp on event syslogger records.
DC syslogger errors themselves now get logged. He has the December
MSDN. He sees a couple requests for scripts to force logoff, so
revises SelfMaint and considers collapsing multiple dialogs into one.
Prior to IAP deployments, Joe will pull KLP out of Pismere.msi. He
will have deployable mirror-distrib, remote event log and ansify by
IAP. He and Asanka work to deploy a new AFS client before the end of
IAP. He will ping Laura May of PSS about W2K3 Server 033 and 039
patch behaviors. He will fix a problem with a CRL user's account, and
study Cygwin binaries and builds.
Team meeting 10 Dec, 2003, 2:00, W92-125
---------------------------------------
Attendees for this meeting are Paul, Richard, Joe, Dave and Tom.
Kakapo meets tomorrow; Joe will represent the team. Brian, Jag and
Robyn all need a team statement on services during closing. Our wash
head breaks due to Jeff's new Kerberos submissions. Paul will get him
onto the build results mailing list, pismere-build. Greg Anderson
requests a client machine for MCC, to which Tom responds.
Paul keeps a list of team member contact information for the closing.
He gladly reports the Tech Time problem is found by Mark and Joe,
requiring an installer fix to work around an AFS directory-creation
limitation. Asanka will code AFS to correct this, perhaps in the
process fixing old problems with IE and SecureCRT APIs.
Richard debugs "Force Local Profiles," sending mail to Laura May. He
creates several Sloan containers and will describe Exchange issues
with them and with Tom Coveney. We discuss BU and their use of a root
E2K server.
Joe makes selfmaint revisions, especially so scripts run under W2K3.
He works with Terminal Services and corresponds to the blat.net
developers.
Dave has OpenAFS tested and checked-in. He has an MSI for LPng and
expects its use by a tester in Milan. He is mapping data flows for
the incremental. He monitors Moira logs and sees no recent errors.
Joe questions whether the incremental-set "associated name" can be
changed, since it may be confused with subcontainer list names.
Team meeting 17 Dec, 2003, 2:00, W92-199
---------------------------------------
Attendees for this meeting are Dave, Joe, Richard, Paul, Qing and Tom.
We review the working to-do list for project wrap-up. See a copy at:
http://web.mit.edu/pismere/meeting-minutes/team/031112.txt
Dave has tested on didnt the EduPerson schema, revised by an Internet2
spec with a new attribute and registry setting. It will go onto isnt
next week. He coordinates with Richard a plan to migrate win over
IAP. He documents AD data flows. The LPng beta tester we expected
unfortunately wants a version for an old Windows OS.
Joe sees MSDN subscriber downloads going away from the site on 23
December. They all rely on the Java VM, and may include some software
we use but do not keep locally. We brainstorm about it and decide the
only liable downloads might be old foreign OSs.
Joe has three development news items. He has a new GP for VS 6 apps
under W2K3S, to be tested on machines in the pismere-* containers. He
has a major revision to selfmaint. Laura May responds to his query
about checksum errors by recommending that autohotfixer should ignore
them.
Richard addresses one main issue - SP4 bugs reported on W2KS is still a bit
mysterious, since PSS mail about LSAs does not answer our question.
This indicates we should test SP4 on our production servers at holiday
break, especially since the Libraries expect to add 100 machines at
IAP.
Paul will head to a meeting with OKI and Mellon to discuss LionShare,
a form of P2P multisite file access with authentication.
Qing has merged OpenAFS with her XML installer for both debug and
normal versions. GNU and Cygwin locker utilities fail on building
Kerberos 5, so she proposes some renaming of the gawk utility. In a
related move, Jeff Altman has new code in the OpenAFS.org build tree.
New data for login statistics reports uses machine, container and
mapping tables on the ttsp database, so we will brainstorm on how to
present it.
January plans
------------
There will be a pair of WIN IAP classes. Container Admins may not see
much new material, but new users should try to attend. See:
win.mit.edu Introduction, Mon, 12 January, 1:00-5:00, 37-312
http://student.mit.edu/searchiap/iap-5455.html
win.mit.edu Advanced, Thu, 15 January, 1:00-5:00, 37-312
http://student.mit.edu/searchiap/iap-5517.html
