Other Peoples Software

1
Other Peoples Software

• Group E
• Interfaces to Scientific Data Archives
• Workshop

2
Background Comments

• Small workshops are great for exchanging ideas
  but not for gathering large samples of data on
  specific packages
• Therefore we had difficulty identifying strengths
  and shortcomings of existing software in areas
  such as data exchange
• need follow-up study?
• see question E-7, for example

3
What software technologies have resulted in
effective user interfaces?

• Html a real success
• object-oriented technologies
• scripting languages
• widgets
• a good example of reusable code
• All of above very broad-based
• used in entire computing industry
• millions use them
• All also very structured

4
What software technologies have resulted in
effective user interfaces?

• The 2 models in The Cathedral and the Bazaar by
  Eric Raymond have both been successful and are
  represented in list above
• bazaar Linux, X, Gnu
• cathedral PowerPoint

5
What software technologies have NOT resulted in
effective user interfaces?

• We punted on this one
• see slide on background comments

6
What software standards have been used in case
studies?
7
What software standards have been used for viz?

• Xlib graphics in XtalView
• great portability across Unix machines, lowest
  common denominator
• obnoxious to maintain and develop but successful
  (users like resulting interface)
• obsolescent, now using OpenGL for graphics
• Xview for GUI
• porting to 64-bit system was difficult

8
What software standards/packages have been used
for viz?

• TCL-tk used in SkyCat (LIGO will probably use as
  its standard)
• interpreted language, similar to LISP
• rapid prototyping
• freely available source for Unix systems
• compiled code available for Suns, SGIs, Windows,
  and Macs
• used in Quanta
• a widely used chemistry package prod. by MSI Inc
• can use the OpenGL API

9
What software standards/packages have been used
for viz?

• LIGO (as did other projects) considered AVS but
  licensing, difficulty of use led to its
  abandonment
• Postscript
• not viz as such but can be used to display
• can be used as an exchange format

10
What software standards/packages have been used
for viz?

• XVT, a comml product (GUI builder)
• portable across Unix, Macs, Windows
• will Java replace it?
• Web-enabled XVT under development

11
What platforms do users want supported for viz
nowadays?

• Windows NT
• Linux
• vendor Unix workstations still major segment
• as opposed to Linux
• Use of web browsers reduces support issues
• but client code may still be platform dependent

12
What Data Exchange Formats Have Been Successful?

• ASCII
• PostScript/PDF
• XDR
• HDF, DIF, etc.
• Gif, jpeg, mpeg,
• sgml

13
What software standards/packages have been used
for DB?

• Nucleic acid DB uses Sybase for relational part
  (but designed so could easily port)
• metadata complicated but schema written in such a
  way that can transform
• ESO has used Sybase since 1985
• as have several others, for some years
• SDSC has used Oracle, not too bad

14
What software standards/packages have been used
for DB?

• Caltech seismo lab is going to try Oracle
• had home-grown till now
• A project started with Britton-Lee Database
  Machine, then ported to Sybase, now going to
  SQL-server (Microsoft, on NT)
• Ecodata project at SDSC now using Oracle for
  metadata interfaced to SRB for HPSS for base data

15
Observations re use of DBMS

• Use DBMS for the metadata, not for the data
  itself if records are very large
• manage the data via pointers
• Use DBMS if have
• fixed data model
• size of records not too large
• have many records
• The point is that metadata easier to move to a
  different DBMS than data itself

16
Observations re use of DBMS

• Relational models provide transportability
• object DBs do not yet have model that provides
  portability
• European Southern Observatory designed interface
  between web and the DBMS to reduce dependence on
  DBMS

17
What software standards are emerging/disappearing
for viz, DB, etc.
18
JAVA or not?

• When will it be portable everywhere?
• There was a report of difficulty in porting to
  Macs, e.g.
• Dependent on a vendors virtual machine
• can cause incompatibility with other packages
• Tower of Babel
• HP introducing its own version,

19
JAVA or not?

• Performance
• Numerical behavior
• NCSA (in HDF context) had difficulty in
  transferring data structures
• Strong typing in Java a plus for portability

20
May (not should) mgt recommend or require that
clients install a particular software package?

• Yes, if
• free or inexpensive
• very easy to install
• runs on most platforms (OS and hardware)
• long-term maintenance/development is feasible and
  planned
• but 15 years is a long time, may not be able to
• counter-example emacs

21
May (not should) mgt recommend or require that
clients install a particular software package?

• Advantages
• avoids having to re-implement many things
• Xlibs, Java beans, e.g.
• higher quality in some cases
• parsing tools Perl, cgi scripts
• numerical algorithms and their implementation
• BLAS 1, 2, 3
• Examples
• MATLAB as required base
• project-developed toolkit as add-on

22
May (not should) mgt recommend or require that
clients install a particular software package?

• In graphics/GUI area, no clear examples
• tend to be tied to underlying system (Motif,
  e.g.)
• But Java will solve everything
• just a joke

23
How protect against changes or disappearance of
software packages, future price gouging, etc.?

• Pragmatic Cant, assume will have to transition
  periodically to other technologies
• High-minded design overall system so that
  swapping subsytems is feasible
• broker paradigm should help

24
When adopting a standard, what is difference
between choosing commercial and free software?

• No marked difference
• support can be as good for free software
• in some cases can buy support for free software
• longevity likewise
• even popular free software can lose support
  (Xview)
• commercial packages of course sometimes die
• Caveat emptor

25
Compare protocols for storing and communicating
data

• In terms of
• portability
• efficiency
• transparency
• openness
• security
• Sidestep issue by using brokers that enable
  separation of storage and transport issues from
  the application
• but then need to choose appropriate broker taking
  5 issues above into account

26
Protocols for storing and communicating data

• There are successful examples, most within a
  discipline
• STAR (self-describing text archive)
• for crystallography, subset called CIF
• in PDS
• ODL (object description language)
• In astronomy
• FITS (flexible image transport)
• HDF
• netCDF

27
What is the transition path from a small to a
large library?

• See the Mythical Man Month
• Plan to re-write the small system to effect the
  transition

28
Gradual learning path for using advanced software
technologies

• Design system so that effort to learn is
  proportional to
• complexity of task or
• level of performance needed
• Analogy using computers with complex
  architectures can ignore architecture if
  performance not an issue
• Eventually will be able to hide complexity

29
How use products of DL Initiative for scientific
archives?

• Need to know more about the products of the DLI
• excuse our ignorance

30
What role is there for CORBA?

• No opinion
• But the broker function is clearly a key one
• for separating applications from specific
  implementations
• for separating data archives from data base
  systems, e.g.

31
When should software be acquired and when should
it be made?

• Build your own
• take care to use the right tools/toolkits
• re-use if at all possible
• using open source software will enable you to
  control your own destiny
• C, C, Java, Perl, tcl-tk will be around

32
When should software be acquired and when should
it be made?

• Acquire packages
• pick software that will probably be around for a
  long time
• use it in a way that provides escape hatch
• be prepared to have to switch (get new product)
• develop a migration plan from the beginning

33
Summary

• As usual with software-related topics, there is
  no clear answer to most questions
• It would be useful to get a larger sample of
  experience data than our small group contained
• There are clear success stories in the use (and
  re-use) of free software and de facto standards
  as well as with commercial packages and formal
  standards

34
Summary

• The use of brokers, of client-server models, of
  web browser interfaces will facilitate the use of
  other peoples software
• Some of the most useful other peoples software
  is in the form of tools (compilers, scripting
  languages) and toolkits (libraries)
• as opposed to complete packages