On Mon, 5 Jan 1998, Mark Wilson wrote:

> Quoting from the publication USMARC Holdings [July 1994, p10], examples for
> fields 853-855, the text for the last example:
>
> 853 20 .........yow05we
>
> appears ambiguous in that it is translated as  "...omits the fifth Wendesday
> of the month."  Although such a reading clearly makes practical sense, it
> appears impossible to distinguish from "omits the fifth Wendesday of the
> year." Even though this latter pattern is highly unlikely, I have to deal
> with this issue because we have implemented pattern prediction in our
> automated serials module and it is very important to defuse possible
> ambiguities [computers, after all, are very stupid].

You are correct in that the way the format appears now, that we have no
way to distinguish between "week of the year" and "week of the month".
The coding for week of the month was added to the format a few years ago
when we did some work on allowing for coding for more complex situations
and exceptions.

> > >From the notes in the publication cited, it appears that to indicate
> unambiguously "omits the fifth Wendesday of each month" requires twelve
> entries in the pattern  "ow0105we" [one for each month of the year].
>
> Conversly, if the example in the USMARC Holdings publication is correct,
> then, how would I deal with a document published each Wendesday  throughout
> the year save for the 5th Wendesday of the year (51 issues)?  Even though
> this pattern is unlikely, I have to tell my software how to distinguish
> between this case and the case cited by the example.
>
> Part of the problem appears to be the interpretation of the patterns MMWWdd
> and MMwwdd. For instance, on the same page:
> MMWWdd     0599tu     Last Tuesday in May
> MMwwdd       0802we    Second Wendesday in August
> note that in this second example, ww has the range 01-53, which would imply
> "Week of the year", not week of the month. Is this an error?
>
> Because there is an overlap in the values for WW and ww (01,02, 03, 04, 05),
> it appears impossible to progam a parser to differentiate any patterns
> containing these overlap values.

There are a few typos on page 9 of the format. Under combinations, under
chronology code definition "w", the pattern wwdd should have been WWdd
(example: 03we Third Wed.). It was intended to be week of month, rather
than week of year. I think the type of publication you mention that is
omitted only the 3rd Wed. of the year was considered unlikely, and noone
apparently thought about it.  In addition the pattern under "w" that now
says MMwwdd should have said: MMWWdd. That error was pointed out to us a
few months ago.  It was intended also to be week of the month.

It seems that the only situations where there could be a conflict would be
the 1st through 5th week of the year (where there is no way to distinguish
between wwdd and WWdd). Again, this situation seems unlikely, but I
suppose we have to allow for all kinds of wierd publication patterns.

It seemed to us that it makes logical sense to use MMWWdd for the week of
the month situation, rather than WWdd. In the case of a publication that
is published (or omitted) the 3rd Wed. of every month, perhaps we could
add codes under Month for "every". We do have codes for "every"(00),
"first" (01), "last" (99), etc. under Week of month. If we also defined
those under Month, the above example for every third Wed. of the month
would be: 0003we instead of 03we.

We have some questions for users of the holdings format if we made this
change:
1) How much use has been made of the existing WWdd pattern in this
subfield?

2) How disruptive would it be to add the codes for every, last, etc. under
month? This would affect the data already coded in $y in such situations.

Note that the regularity codes are maintained by LC, so we just need to
announce changes and incorporate them into the format as needed.

> > Or is there some subtext in the use of WW and ww that I have missed?
>
> Finally, how do I indicate a publication that is not published on any
> holiday, including movable feasts?

Aren't holidays on predictable days, like the last Thurs. in Nov. for
Thanksgiving and the 3rd Mon. in Feb. for President's Day? Others
are actual days, i.e. Dec. 25th.  Couldn't the codes already
established be used for those?

> Thanks, and advice will be appreciated.
> Mark
>
> Mark Wilson, Director R&D
> The Library Corporation
> Research Park, Inwood, WV 25428
> 304-229-7803 (Voice)
> 304-229-0295(FAX)
> Solutions that deliver!
>

We would appreciate any feedback, especially from those using 853-855$y
for predicting publication patterns.

********************************************
Rebecca S. Guenther
MARC Standards Specialist
Network Development & MARC Standards Office
Library of Congress
Washington, D.C. 20540-4102
Phone: 202-707-5092; FAX: 202-707-0115
[log in to unmask]
********************************************