1/LOW/2009-01-27/FIXED
	Naming in protocols/scs.py -- SCS should be consistently upper case

2/MEDIUM/2008-10-20/WONTFIX
	StreamWriter errors cannot be propagated up yet. [Error handling must
	really be in the writing thread, since the calling function cannot know
	if anything has been written yet.  Of course, when things go wrong
	between the handover and the start of stream writing, there may
	be empty output, but this needs to be handled some other way]

3/WISHLIST/2009-01-10/FIXED
	gavoimp should have some functionality to delete everything described
	by an RD [gavodrop does this]

4/MEDIUM/2009-01-24/FIXED
	Remove imageFormat column in SIAP mixin and take the data from
	product's mime.

5/MEDIUM/2009-01-27/FIXED
	formalTypes for InputKeys should be computed at form construction time
	to fix the mess with required condDescs having non-required InputKeys.

6/WISHLIST/2009-01-27/FIXED
	The sidebar should become a template on disk.

7/LOW/2009-02-02/FIXED
	MultiSelections are not handled correctly when used in bookmarkable forms 
	-- only the first selection is kept.

8/WISHLIST/2009-02-10
	Columns, Tables, Data Descriptors should get stripped-down base classes,
	such that an OutputTable doesn't have an adql attribute.  It simply looks
	bad in the docs.

9/LOW/2009-02-11/FIXED
	Unify credentials checking in services (runAuthenticated) and products.
	Probably both should be using the Authenticate exception. [has been
	fixed for quite a while before revision 1466]

10/LOW/2009-02-11/FIXED
	Turn UnicodeAttributes in grammars/kvgrammar.py to REAttributes as 
	appropriate.

11/MEDIUM/2009-02-11/FIXED
	Documentation for row triggers currently doesn't really exist.

12/MEDIUM/2009-02-11/FIXED
	Julian year functions are wrong.  Use 
	datetime.datetime(2000, 1, 1, 12)+datetime.timedelta(days=365.25*(fracYear-2000.0) 
	(which still is not 100% precise but should really be good enough).

13/MEDIUM/2009-02-14/FIXED
	SELECT 'foo' FROM whatever raises a weird error

14/MEDIUM/2009-02-14/FIXED
	SELECT 1 as size is not parsed.

15/MEDIUM/2009-02-18/FIXED
	The whole formalType/inputWidget thing with InputKeys is an absolute
	pain.  At least there should be way to suppress or override the automatic 
	vizier type conversion. [using formalType]

16/MEDIUM/2009-04-17/FIXED
	Conversion between from TAI to UTC computes location of leap seconds based
	on TAI rather than UTC.  This will lead to incorrect UTCs in the vicinity
	of the leap seconds.

17/MEDIUM/2009-04-20
  Pre-1972 TTs should probably be interpreted as ET and converted to GMT
	in TTtoUTC -- but they are not.
	See http://hpiers.obspm.fr/eop-pc/earthor/utc/TAI-UTC_tab.html

18/LOW/2009-06-29/FIXED
	user.mkrd still contains pre-rewrite code and thus doesn't work.

19/LOW/2009-06-30
	Naming of the various data structures that deal with column properties
	in formats.votable is a complete nightmare.  The collection of column
	properties for a table needs a catchy name.

20/WISHLIST/2009-07-14
	The whole customwidget/gavo.js stuff is a mess.  Do output format, DB, etc.,
	without formal widgets.  This stuff is better done manually.  Actually
	create all necessary widgets in the result tree and scrap the dynamic
	building from gavo.js.

21/MEDIUM/2009-07-17/FIXED
	Bad profiles (e.g., no database set) will cause getDBConnection to bomb
	out in an ugly way.  It should at least provide some sensible information.

22/SEVERE/2009-07-17/WONTFIX
	The creation of metaHandler in rscdesc leads to a deadlock when 
	dc_tables does not exist yet. ["So don't do that" -- this has not
	been necessary for quite a while]

23/MEDIUM/2009-07-21/WONTFIX
	Conversion of unknown VOTable types leads to None with FromVOTableConverter.
	It should yield an exception. [Code doesn't exist any more]

24/SEVERE/2009-07-21/FIXED
	A Cone Search of 180 degrees around 10,10 doesn't return the whole table
	(actually, it apparently returns nothing) [was a q3c problem]

25/SEVERE/2009-07-21/FIXED
	Linefeeds embedded in queries coming from the form cause the service
	to bomb out (actually, it was only trailing whitespace)

26/MEDIUM/2009-07-30/FIXED
	binarize on STC ASTs with compounds doesn't work any more.

27/WISHLIST/2009-08-04/FIXED
	It would probably be helpful if instanciating an RD with a non existing
	resdir would at least issue a warning of some sort.

28/LOW/2009-08-04
	"AS" clauses should override any names obtained from system tables in ADQL.
	And we should be more careful inheriting such properties through
	functions (e.g., SQRT should probably kill almost all such properties).

29/LOW/2009-08-31
	structure.EventProcessor doesn't update root element when it has been
	Replace'd (proposed solution: Kill the abominable Replace)

30/WISHLIST/2009-09-02
	meta.ModelBasedBuilder is a horrible abomination.  Replace it with
	something that allows specification of required elements and a sane
	syntax, or, probably better, teach xmlstan the render trick of the real
	stan and write renders for meta items/meta values.

31/MEDIUM/2009-09-04
	STC information is missing in the Box mapper in valuemappers.  It's not
	always ICRS like we're claiming now.

32/MEDIUM/2009-09-08/WONTFIX
	Form values in split-up date fields are not preserved. [Actually, they
	are, if all items are given; partial values are very hard to preserve]

33/MEDIUM/2009-10-02
	We do not actually give availability information so far.  Cores should have
	some standard way to answer availability questions, and the main system
	should be able to override what the service says.

34/LOW/2009-10-05/FIXED
	Replace RDNotFound exception with the more generic NotFoundError.

35/MEDIUM/2009-10-08
	Non-required defaulted fields (like cutout size on lswscans/positions)
	do not get their default filled in on submitted forms.

36/LOW/2009-10-16
	base.__init__ calls pyparsing...setDefaultWhitespaceChars.  It really
	shouldn't do this.  Find out what grammar needs this and do a local
	setDefaultWhitespaceChars or similar.  Then remove the setDefault... calls in
	adql.getGrammar.

37/HIGH/2009-10-19/FIXED
	ROUND and TRUNCATE should accept a second argument in ADQL.

38/MEDIUM/2009-11-03
	You can select nonsensical (on-selected) fields in the sort widget.

39/LOW/2009-11-16
	Services with verbLevel-defined outputTables are misclassified as
	TableServices if they contain positional data.  That's kinda hard
	to fix except by delaying service type determination.

40/HIGH/2009-12-11/FIXED
	select max(ntotal), min(ntotal) from usnob.platecorrs only returns one 
	column.

41/MEDIUM/2009-12-21/FIXED
	When vo_managed-publishing something, try producing a resource record and
	complain if that doesn't work out. [this has been fixed a while before
	revision 1466]

42/MEDIUM/2010-01-25
	Try condDesc original="ref" referring to a column; this results in a
	completely unintelligble error message.

43/HIGH/2010-02-25/FIXED
	rave.main and tspectra.data have no descriptions on 
	__system__/dc_tables/list/form -- why?

44/WISHLIST/2010-04-10
	Make a "noform" renderer (or so) from the runOnEmptyInputs case of
	resourcebased.form

45/MEDIUM/2010-04-13
	Fields with formatters are not included in VOTables at all?

46/LOW/2010-07-26
	UWS job directory is not created if it does not exist yet.

47/HIGH/2010-07-30/FIXED (rev. 1635)
	See input/glots: The createView data is touched but has no effect when
	doing createTables.

48/LOW/2010-08-26
	taprender should return http 404 code rather than 400 for a NotFoundError

49/MEDIUM/2010-09-02
	No notes on service info pages

50/HIGH/2010-09-02/FIXED
	select ab || 'cd' from x doesn't parse in adql (valueExpression needs 
	to reconsider if numericValueExpression parses but bad things happen
	downstring).

51/MEDIUM/2010-09-06
	N-dimensional arrays basically don't work in votable for N>1.  Even for
	N=1, it's not pretty.

52/LOW/2010-09-09
	The argparse framework in user (used, e.g., in gavo admin) requires
	subcommands to be spelt out.  So, you can say gavo ad listusers but
	not gavo ad listu.

53/MEDIUM/2010-09-27/FIXED
	Have grammars be referred to by name rather than as object from outside
	the grammar sub-package.  Then, stop preloading all grammars.

54/MEDIUM/2010-09-27
	Do away with all the "predefined" mess with procs and such.  They
	should be referred to by references (possibly allowing skipping
	__system__).  Then, do away with the preloading crap for mixin RDs.

55/WISHLIST/2010-10-07/FIXED
	A few utility functions from votablegrammar should be put somewhere else
	(valuemapper?) so taprunner, formats.votable* and friends don't need to
	import it any more on startup.

56/LOW/2010-10-08/FIXED
	Kill the ref attribute and the underlying mechanism for structures.
	It has been a pain and not terribly useful.

57/MEDIUM/2010-12-10
	SIAP services should compute their sia.type from the core they have.

58/HIGH/2010-12-10/FIXED
	pyparsing is not thread-safe.  Protect all parses by locks (probably
	the lock objects should sit on the grammars, and I should have a
	common parse method). [more or less fixed, some time around rev. 1700]

59/MEDIUM/2011-01-25/FIXED
	ADQL count(*) returns a float. [in ref. 1788]

60/WISHLIST/2011-01-25
	When using utils.intToFunnyWord to make up ids, we should use some
	reserved marker to avoid collisions with user-defined ids.

61/WISHLIST/2011-02-18
	Do away with VOColDesc and instead add some additional data to the
	column objects themselves.

62/MEDIUM/2011-02-23/FIXED
	http://dc.zah.uni-heidelberg.de/rauchspectra/q/tovot/info has a URL
	for a service that has no web frontend. [fixed ca. rev. 1800]

63/MEDIUM/2011-03-09
	gavo imp --system //uws does not remove stale uws tables.  Why?

64/WISHLIST/2011-03-15/FIXED
	Generate prettier names for ADQL result tables. [in rev. 1816]

65/WISHLIST/2011-04-04
	SIAP tables should have their dates in MJD; that helps obscore a lot
	and may obsolete some ugly hacks in valuemapper.  Wenn doing this, don't
	forget to fix obscore#publishSIAP.  While doing that, require exposure
	time in SIAP and fix tMin and tMax.

66/MEDIUM/2011-04-06/FIXED
	ADQL INTERSECTS with a point is supposed to be an overlap.  In our
	current code, it's an error. [fixed somewhere around rev. 1900]

67/MEDIUM/2011-06-07
	VOTable PARAM with an NULL (None) value just gets to be a value="" attribute.
	That's bad, we should use VALUES to define null literals.
