Proposal: fixed field order

[Ka-Ping Yee]

PFIF 1.1 specified a fixed order for the fields in each record.

PFIF 1.2 relaxed this, requiring certain fields to be first and then allowing the rest to be in any order.

I am now feeling that this was a stupid mistake on my part, for two reasons:

1. The field order constraint in PFIF 1.2 cannot be expressed in XSD, so PFIF 1.2 cannot be validated against an XML Schema.

2. It is easier to work with conversions to/from CSV, spreadsheets, or other formats if the fields are in a fixed order.

Therefore, I propose that PFIF 1.3 return to specifying a fixed field order, consisting of the fields in PFIF 1.1, followed by the new fields in PFIF 1.2, followed by any other new fields.

Any feedback on this?

—Ping
Google Crisis Response

[Peter Sutton]

What motivated the move from the fixed field ordering of PFIF1.1 to
the relaxed field ordering of PFIF1.2?

Thanks,

Pete.

(I had a quick look your documentation but I didn't anything. If I
missed it, I'm sorry.)

[Ka-Ping Yee]

On Fri, Oct 22, 2010 at 05:11, Peter Sutton <petersu...@gmail.com> wrote:

What motivated the move from the fixed field ordering of PFIF1.1 to
the relaxed field ordering of PFIF1.2?

It seemed nice to be able to make up a PFIF file by hand and have it be valid, without having to memorize the field order.  Or to write scripts that emit PFIF without worrying about field order.  It is kind of a pain to get right.

I'm curious to hear from other implementors on this—do existing implementations get the field order constraints in PFIF 1.2 right?  If not, that might be a sign that we should relax it.

On the other side of the issue, how valuable is it to have an XSD at all?  Is anyone trying to validate?  It is interesting that other major formats, such as Atom itself, allow fields in any order, and no one seems to be suffering for lack of an Atom XSD.

Note that though XSD has trouble allowing fields in any order, RELAX-NG can do it:

    http://www.relaxng.org/compact-tutorial-20030326.html#id2815185

Some feedback would help us evaluate both sides of this decision.

—Ping