After considering the question of how to best represent names in a database in a way that's both accurate and neutral to local naming forms, the best way is to avoid terms like "first name", "last name", "given name", "family name", "surname", etc. If you use terms like this there will always be some exception to your rule.

Therefore the terminology that I prefer to use is "sort name." In other words, if you want to find Hou Yifan in the Player Directory, do you go to the "H" or the "Y"?

This terminology solves a lot of pedantic quibbles since it gets to the meat of why we even care in the first place. The software doesn't even know what a "first name" or "last name" is, nor should it, it just needs to know what to put on the computer screen in the various contexts.

The end users have many desires depending on the context. On a player page most people want to see names represented as accurately as possible, but when searching they don't want to be accurate whatsoever. (By the way, Penguincw: just search for "Nepo" and it will work.)

In some situations, like game lists, users want to see names collapsed as compactly as possible. In other circumstances, like the player page, it is nice to have access to the most expanded and diacritically accurate version.

One possible schema for representing names that could meet these requirements would be to provide four (!) different strings for each player, some of which are optional. For example:

FULL NAME: Peter Leko
SHORT NAME: Leko
REGIONAL NAME: Lékó Péter
SORT NAME: Leko

or in the case being discussed,

But please use the full name, adopting FIDE comma convention.

And use the <sort names> in the PGN.

Ultimately the rules will be based on how we fill out the database; because any rule that assume "Asian names work like this ..., Spanish names work like this ..., etc." is doomed to fail. So there will be a time when some players are working fine and others are still unusual, but we can migrate toward perfection.

I don't think so...

E.g. how about separating all the <Petrosian>'s in the list?

Or <Rodriguez>'s, or <Smith>'s, etc., etc.

Are you really going to do a sort twice, first on the sort name, and then on the full name?

If you use the FIDE convention, you don't need to store two names in the DB, the "sort name" is just fine and dandy.

There is then a very simple rule to convert the "sort name" to the "display name" -

Just put the part after the comma in front.

If no comma, then "sort name" = "display name".

If you happen upon the player in person, address them formally with the pre-comma part of the name and you'll be right, or close to correct.

Publish the PGN with the "sort name" by default.

If anybody prefers the old-style PGN, have a user flag to accommodate them.

But let's get the default correct, and streamline the DB accordingly. And keep the "rules of engagement" as simple as possible.

The great advantage of this convention is we get to leverage the work of FIDE, while conforming to virtually all other chess DB's in existence.

Are you really going to do a sort twice, first on the sort name, and then on the full name?>

I'm not sure I follow. What do you mean by separate?

Sorting on two fields is not the same as sorting twice. The player would be indexed on the combined fields (SORT NAME, FULL NAME). Then a list like this

Tigran Vartanovich Petrosian (sort = Petrosian)
Arshak B Petrosian (sort = Petrosian)
Yifan Hou (sort = Hou)
Tan Zhongyi (sort = Tan)
Viswanathan Anand (sort = Anand)

would come out like this:

Viswanathan Anand
Yifan Hou
Arshak B Petrosian
Tigran Vartanovich Petrosian
Tan Zhongyi

or if you prefer

Anand, Viswanathan
Hou, Yifan
Petrosian, Arshak B
Petrosian, Tigran Vartanovich
Tan, Zhongyi

(Note in the above example I kept Yifan's name in the "wrong" order, and used Anand as an example of where there is no logical reason why it should be in that order other than "he says so.")

   Carlos Repetto Torre

sometimes is written

   Carlos Torre Repetto.

I'm pretty sure the first form (as we have it) is the correct rendering, but for the sake of argument let's see how the proposal can handle both cases. If <Carlos Repetto Torre> is correct then we have

FULL NAME = Carlos Repetto Torre
SORT NAME = Torre
DISPLAYS LIKE THIS: "Torre, Carlos Repetto"

If <Carlos Torre Repetto> is regarded as correct then we have

FULL NAME = Carlos Torre Repetto
SORT NAME = Torre Repetto
DISPLAYS LIKE THIS: "Torre Repetto, Carlos"

Q - Why not have just the one name, the "sort-name", in FIDE-like fashion?

It saves space - eliminating needless duplicate.

It saves time, and improves performance - by only having to sort on one field.

It allows a very simple algorithm to be used to convert from "sort-name" to "full-name".

* * * * *

So, suppose a FIDE name is <C D, A B>

The full name displayed on the player page is

<A B C D> (simple rule: swap after-comma and before-comma)

The name in the PGN used is the "sort-name", just like TWIC, CB, MillBase, Carolus, Weeks, etc. etc. all use.

<I'm not sure I follow. What do you mean by separate?>

OK, let's look at a search on Petrosian, here's the result:

<

Tigran Levonovich Petrosian (2671) 592 1994-2017 Tigran Vartanovich Petrosian (2645) 1921 1942-1983 Arshak B Petrosian (2510) 233 1970-1999 Davit Gevorgi Petrosian (2504) 95 2001-2016 Suren Petrosian (2408) 30 1980-2016 Tigran A Petrosian (2383) 12 2001-2004 Karen Petrosian (2350) 6 1993-1996 Karlen Petrosian (2348) 2 2006-2010 >

By separate, I meant sort - as in - how do I distinguish from all the Petrosians above?

The list <CG> yields is sorted by rating. But the more generally useful sorting would be by "sort-name".

(E.g., historical players don't have a rating. In fact, nobody before ~1970 does. Also, given rating inflation - sorting by rating won't always put the strongest player at the top.)

Of course, ideally, it would be nice to sort by clicking on column headings.

Let's assume that we do want to sort by sort-name. Then we want to display the sort-name to guide the eye.

It's much easier to scan a column of names. I constantly stain whenever a search yields 5 or more names.

Consider a search for Rodriguez:

http://www.chessgames.com/perl/ches...

The reason? Well, the years aren't lined up in a column.

Or rather, some entries have just one year, others have two (start-finish).

This makes it harder on the eye, then if the start years were all aligned (and the finish year either left blank or given).

The name is <Torre Repetto, Carlos>.

http://www.chessgames.com/perl/ches...

A complete jumble of the wanted and unwanted.

It would be nice to just list the sort-names which match the length of the search term.

One can imagine introducing the use of simple wildcards to allow searching for a first name.

The player list is not complete - only a portion of players are there.

(Switch knows the actual percentage, I could calculate with some effort but the idea is to have 100% coverage)

Second, the playerlist format isn't used anyplace else but there.

Third, it's not used in the PGN.

* * * * *

I would think that the "full-name" only belongs at the heading of the bio page.

All search list results should use the "sort-name".

* * * * *

I wasn't kidding about some names being so "full" as to be cumbersome.

Here's another one:

Some have been used, but not all.

A litmus test case I'm particularly interested in is Junge.

Do you intend to publish the unrestricted PD photograph I sent to you?

I had a look at his bio. Apparently he died February 23, although it was also reported as February 24. Apparently the uncertainty of that caused the following problem. The spot on the bottom of his player page for a bio writer to list DoD has:

Date deceased: 2017.02.00

This shows up in the bio as:

ENVER BUKIC

(born Dec-02-1937, died Feb-??-17, 79 years old)