biblScope should be in Imprint or not? (biblStruct)

[TEITechnicalCouncil]

In the Guidelines, 3.11.2.3 says:

"A bibliographic description, particularly for an analytic title, will often include some additional information specifying its location, for example as a volume number, page number, range of page numbers, or name or number of a subdivision of the host work. The element biblScope may be used to identify such information if it is present. Where it is desired to distinguish different classes of such information (volume number, page number, chapter number, etc.), the type attribute may be used with any convenient typology.

"When the item being cited is a journal article, the imprint element describing the issue in which it appeared may contain biblScope elements for volume and page numbers, together with a date element."

I believe the first paragraph explains how to use a <biblScope> that is not a child of <imprint>, and the second paragraph describes a <biblScope> which is a child of <imprint>. What makes the description of a journal article different from the description of an analytic title? I thought a journal article's title is a type of analytic title.

So there are a few things that need clarification:

1) Whether the first paragraph's directions about use of biblScope really means to say that it is a sibling, not child, of imprint.

2) Why a journal article is different from an analytic title (or other analytic titles).

3) Whether, in the case of a journal article, there should be a biblScope sibling of imprint in addition to a child. It seems that the sibling biblScope would have the page numbers cited and the child biblScope would have the page numbers of the whole article.

Original comment by: @kshawkin

[TEITechnicalCouncil]

Maybe we should add an example that would illustrate this difference, as follows: If one happens to give the following as a reference, we agree it means the book chapter appear between pages 347 and 374 of the edited book: <biblStruct type="inbook" xml:id="inria-00075838"> <analytic> <author> <persName> <forename>Serge</forename> <surname>Abiteboul</surname> </persName> </author> <author> <persName> <forename>Stephane</forename> <surname>Grumbach</surname> </persName> </author> <title type="main" level="a">COL : a logic-based language for complex objects</title> </analytic> <monogr> <title level="m" type="main">Advances in database programming languages</title> <imprint> <publisher>ACM Press</publisher> <biblScope type="publicationDate">1987</biblScope> <biblScope type="fpage">347</biblScope> <biblScope type="lpage">374</biblScope> </imprint> </monogr> <idno type="doi">10.1145/101620.101641</idno> <idno type="url">http://hal.inria.fr/inria-00075838/en/&lt;/idno&gt; </biblStruct>

And if one adds page information like here, it means one quotes page 349:

<biblStruct type="inbook" xml:id="inria-00075838"> <analytic> <author> <persName> <forename>Serge</forename> <surname>Abiteboul</surname> </persName> </author> <author> <persName> <forename>Stephane</forename> <surname>Grumbach</surname> </persName> </author> <title type="main" level="a">COL : a logic-based language for complex objects</title> </analytic> <monogr> <title level="m" type="main">Advances in database programming languages</title> <imprint> <publisher>ACM Press</publisher> <biblScope type="publicationDate">1987</biblScope> <biblScope type="fpage">347</biblScope> <biblScope type="lpage">374</biblScope> </imprint> <biblScope type="pp">349</biblScope> </monogr> <idno type="doi">10.1145/101620.101641</idno> <idno type="url">http://hal.inria.fr/inria-00075838/en/&lt;/idno&gt; </biblStruct>

Original comment by: @laurentromary

[TEITechnicalCouncil]

There seem to be several quite different issues here.

1. The first paragraph Kevin cites should not be read as implying anything about where exactly a biblScope is located within the biblStruct in question. It does not say anything about the topic, but merely clarifies that additional information may be conveyed by that element.

2. The difference between a journal article and a different kind of analytic article is that the former appears in a journal (i.e. a serial of some sort). Otherwise there is no difference, vide discussion on the TEI-list last month about Balzac!

3. In the case of a journal article (or any other analytic title come to think of it), the biblScope explaining where the article (or whatever) occurs within its monographic parent item can be located anywhere within that item. In the case of a serial, the most convenient place is usually the imprint, but it could also be as a sibling. I am not convinced of the wisdom of giving this distinction additional significance, see below.

4. In my view, the requirement to be able to say "this is a reference to page4 of an item which occupies pages 3 to 6" is quite a different issue. If there were four different citations to four different places in the same article you wouldn't want to put the full biblstruct for the article in your bibliography four times, surely? I think the way to do that is to use the <ref> element. <ref target="#foo">p 4</ref> <ref target="#foo">p 5</ref> where #foo points to a <bib;lStruct> for the whole article with a biblstruct pp 3 to 6.

Original comment by: @lb42

[TEITechnicalCouncil]

• milestone: --> 871214

Original comment by: @lb42

[TEITechnicalCouncil]

Responding to Lou's comment ...

1) What's confusing here is that the first and second paragraphs that I cite discuss use of biblScope for volume and page numbers. It seems like some sort of opposition is being established depending on whether the cited work is a journal article. If the Guidelines do not mean to prescribe a difference here, these two paragraphs need to be reworded so as not to discuss volume and page numbers twice.

2) I can't find the discussion on Balzac. Please provide a pointer.

3) Prescribing a difference would aid in machine processing of citations. For example, OpenURL resolvers depend on a specific citation structure, so if we don't prescribe one way or the other, even a citation as structured as biblStruct won't be machine-actionable.

4) Your example using ref elements describes how you'd encode references to footnotes/endnotes (the little inline superscript numbers), not the endnotes themselves. I agree on how you'd encode the references, but you said "you wouldn't want to put the full biblstruct for the article in your bibliography four times, surely". I'm not so sure of this: if I were encoding footnote references for easy machine processing by an OpenURL resolver, I would in fact repeat the biblStruct over and over, though I admit this wouldn't work well if you were trying to represent a print source document that used an abbreviation like "Ibid."

In any case, I see this as a separate matter. The question at hand is whether the TEI Guidelines should recommend anything specific regarding placement of biblScope when there's an imprint present. It would also be helpful to say something about what to do when you have two page number (or page ranges) given: one for the part cited and one for the item at the analytical level.

Original comment by: nobody

[TEITechnicalCouncil]

Just noticed that I neglected to log in before posting comment #3 (the previous one, from 2009-06-01 16:53).

Original comment by: @kshawkin

[TEITechnicalCouncil]

Looking at this ticket with fresh eyes, let me summarize the issues.

1) §3.11.2.3 says:

"The elements biblScope, pubPlace and publisher constitute the special class model.imprintPart; members of this class may appear with a date inside an imprint element in a specific location within a biblStruct, or alternatively, they may appear alongside any other bibliographic component inside a bibl."

Indeed, all examples given in §3.11.2 show biblScope as a child of imprint.

However, §3.11.2.7 describes use of biblScope as a child (not descendant) of monogr, located in a particular location (contradicting Lou's comment from 2009-06-01 16:22). Furthermore, according to the element definition, biblScope may be a child of:

monogr series imprint bibl

Furthermore, Laurent's comment from 2009-03-27 06:23 shows a good use case where you would want to allow both types of biblScopes, but Lou suggests in his comment from 2009-06-01 16:22 that you would instead do this with ref elements. It' not clear whether Lou's proposed method of encoding was meant to be used within a biblStruct/biblFull that describes something like "page 4 of an item which occupied pages 3 to 6" or whether it was meant to encode the footnote in the text.

2) §3.11.2.3 has a paragraph beginning "When the item being cited is a journal article". It seems that this is an elaboration of what's described in the previous paragraph (about analytic titles), and therefore it should be the last sentence of the previous paragraph. As it is, it appears that journal articles are different from other analytic titles. An alternative way of resolving this is to insert "In particular, " before "when the item being cited is a journal article".

Original comment by: @kshawkin

[TEITechnicalCouncil]

On balance Council agrees that biblScope should be a sibling rather than a child of imprint, but it is recognised that there are several related issues that need to be teased out further before we can resolve this.

Original comment by: @lb42

[TEITechnicalCouncil]

In between the 2009-12-07 and 2010-02-08 Council conf calls, Kevin, Laurent, and Lou had the following discussion:

<blockquote> As for (1), I agree with Lou that

"page 367 of [title] which occupies pages 145 to 678 of ...."

is different from

"[title] which occupies pages 145 to 678 of ...."

Laurent's comment from 2009-03-27 on the ticket shows that he wants to encode these things differently, and I do too (though didn't state so explicitly when opening the ticket). Laurent and I both assumed this would be done using the biblScope element in different locations, but Lou says he wants to use ref for "page 367 of [title] which occupies pages 145 to 678 of ...."

Lou, could you provide an example of encoding, preferably using these same page number for clarity? You've made broad statements about how to do this (on the SF ticket and in your most recent email), the details are hazy to me.

As for (2), Lou suggests allowing biblScope only as a child of monogr or series (not as a child of imprint) and suggests specifying that the content of biblScope describe the pagination etc. of the bibliographic item in the sibling title element. A few questions here:

a) Since you would no longer be able to encode the 367 of "page 367 of [title] which occupies pages 145 to 678 of ...." using biblScope, we would need another mechanism here. Are you still suggesting use of ref for this?

b) bibl does not require use of analytic, monogr, or series as child elements. Instead of requiring that biblScope have monogr or series as a parent, wouldn't it be better to forbid use of biblScope inside of imprint? Or are you explicitly trying to block people from putting biblScope inside of analytic?

--K.

On 28/01/2010 17:32, Lou Burnard wrote: > > We have two issues, I suggest: > > 1. Is a reference like "page 3 of [title] which occupies pages 145 to > 678 of ...." > the same kind of animal as "[title] which occupies pages 145 to 678 of > ...."? > > I feel that it is not, hence my suggestion that it should be treated as > a <ref> rather than a <biblStruct>. But I can see why people might > disagree and I don;t think it matters that much(you can always wrap your > <ref> in a <bibl> if you like) > > > 2. More problematic, is the possible location/s of <biblScope> within > <biblStruct>. Reviewing the examples (for <imprint>) in the Guidelines, > it's clear that we're not completely inconsistent about whether > <biblScope> should appear as a child or as a sibling of it -- this is > partly because I did try to make it consistently a child, rather than a > sibling, the last time we discussed this issue, but some examples got > away (e.g. Nelson80) ... but in any case, I now feel this was a mistake > and that <imprint> should contain information only "relating to the > publication or distribution of a bibliographic item" i.e. who dunnit, > when and where. I don't think the "where" justifies adding information > about which volume to look in to find the item concerned. Note that we > also have <extent> information ("three volumes") which is a sibling, not > a child , of <imprint>. So the parent for <biblScope> should always be a > <monogr> or a <series>, and it should always be a sibling of the <title> > (at whatever level) for the item being documented. > > > >> Hi there, >> I have read this again (probably the 5th time) and I see a conceptual >> stumbling block on the issue of "what does a bibliographical reference >> refer to". One this is solved, we may just find an easy solution to >> the biblScope question. Let me try to express one (possible) semantics >> for <biblStruct> and see the possible consequences: >> We could see biblStruct as providing a piece of information containing >> both descriptive and locative data for it (yes, I know this all about >> metadata....). The locative information is there to allow someone to >> retrieve the actual piece of information, possibly in its context. >> If one wants to refer to a full paper, the situation is easy: the >> biblScope in <imprint> give the place (in relation to volume, issue >> information) of this same paper. >> If one want to refer to a specific page, I would say, this is exctly >> the same, the biblScope in <imprint> should contain the page information. >> All in all, this means I was wring when trying to have two >> <biblScope>'s at two different places. It may be the case that we >> could "impose" that <biblScope> should only occur in <imprint> and >> provide a clear theory of reference for this. >> Am I getting stupid? >> Laurent >> >> >> >> Le 27 janv. 10 à 20:17, Kevin Hawkins a écrit : >> >>> Lou? >>> >>> On 08/12/2009 14:55, Kevin Hawkins wrote: >>>> To make sure we're all not waiting on each other, I believe Lou >>>> needs to >>>> take the next step with this ticket: >>>> >>>> https://sourceforge.net/tracker/index.php?func=detail&aid=2714682&group\_id=106328&atid=644062 >>>> >>>> >>>> >>>> He'll look at the prose revision in (2). As for (1), he'll give some >>>> examples of how the content of biblScope gives "scope" to the parent >>>> element. This would require revision of Laurent's examples to >>>> demonstrate to Laurent and me the proper use of biblScope. Once we can >>>> agree on the proper use, we'll figure out how to best revise the prose >>>> in §3.11.2 to make it clearer how to use this element. >>>> >>>> Sound right? >>>> >>>> Kevin >>>> >> > </blockquote>

Original comment by: @kshawkin

[TEITechnicalCouncil]

To be addressed in context of several other bibliographic issues addressed by draft from Kevin, Martin, Laurent.

Original comment by: @lb42

[TEITechnicalCouncil]

This initially addresses some vagueness in the guidelines with regard to the use of <biblScope>: when should it be a sibling of <imprint>, and when should it be inside <imprint>? However, the discussion has broadened so much that we could now say it encompasses these questions:

1. What does <biblScope> mean when it is a child of <monogr>, as opposed to a child of <imprint>?

2. Should <biblScope> be removed as a child of <imprint>?

3. Should <biblScope> be allowed as a child of <analytic>? If not, why not?

I have no real view on #1, but I have a feeling that it will be impossible to achieve any consensus on it.

I say an emphatic NO to #2, because it breaks backwards compatibility for no appreciable gain.

I say YES to #3, because in my view it would allow more logical encoding (page numbers of a chapter arguably pertain to the chapter rather than the book that contains it).

Kevin has responded to this comment by email, and will post his comment here.

Original comment by: @martindholmes

[TEITechnicalCouncil]

In our Dublin proposal, we settled on the following:

monogr/biblScope: the cited page(s) of an article monogr/imprint/biblScope: the starting and ending page of an article

I think this would stand even if we give up <biblScope type="fpage"> and <biblScope type="lpage"? in favor of <biblScope type="pp" from="foo" to="bar">.

That is, we are choosing (1) and clarifying current practice.

Original comment by: @kshawkin

[TEITechnicalCouncil]

The biblio gang (Kevin H, Laurent R and Martin H) recommend that this ticket be set aside for the moment; it represents a set of questions rather than a straightforward request. We propose to address the issues raised in this ticket in more detail in our upcoming biblio revision work.

Original comment by: @martindholmes

[TEITechnicalCouncil]

• status: open --> open-postponed

Original comment by: @lb42

[TEITechnicalCouncil]

• status: open-postponed --> pending-postponed

Original comment by: @lb42

[TEITechnicalCouncil]

This Tracker item was closed automatically by the system. It was previously set to a Pending status, and the original submitter did not respond within 14 days (the time period specified by the administrator of this Tracker).

Original comment by: sf-robot

[TEITechnicalCouncil]

• status: pending-postponed --> closed-postponed

Original comment by: sf-robot

[TEITechnicalCouncil]

• milestone: 871214 --> RED
• status: closed-postponed --> open-postponed

Original comment by: @lb42

[TEITechnicalCouncil]

• status: open-postponed --> open-accepted

Original comment by: @kshawkin

[TEITechnicalCouncil]

It seems that we want to recommend the following tagging practice:

monogr/biblScope: the cited page(s) of an article whose title is given in analytic monogr/imprint/biblScope: the starting and ending page of an article whose title is given in analytic

So we should make the following changes to P5:

1. In "NELSON80" biblStruct example, move <biblScope type="pp">1013–23</biblScope> into the imprint element just above it.

2. In PMLA example, change <biblScope type="pages">122</biblScope> to <biblScope type="pp">121-139</biblScope> and insert <biblScope type="pp">122</biblScope> after </imprint>. (The value of @type is the practice prescribed in section 3.11.4 and used elsewhere, so we should fix it while we're making other adjustments to this example. As for having more than one biblScope, this citation is a good example of the two types given above.)

3. In this sentence:

The elements biblScope, pubPlace and publisher constitute the special class model.imprintPart; members of this class may appear with a date inside an imprint element in a specific location within a biblStruct, or alternatively, they may appear alongside any other bibliographic component inside a bibl.

insert "or biblFull" after "within a biblStruct". Then, after the above sentence, add the following:

In biblStruct or biblFull, the element biblScope may appear as a sibling or child of imprint. Use of each is described below.

4. Revise this sentence:

When the item being cited is a journal article, the imprint element describing the issue in which it appeared may contain biblScope elements for volume and page numbers, together with a date element.

as follows:

When biblStruct or biblFull is used, and when the item being cited is a journal article, the imprint element describing the issue in which it appeared may contain biblScope elements for volume and page numbers, together with a date element, indicating the location of the analytic item. But if biblScope is a sibling of imprint, it indicates particular volume or page numbers cited by the author.

5. Revise the "Parish registers and the historian" example to include <biblScope type="pp">161</biblScope> after </imprint> and add the following sentence after the example:

In the above example, the article is found on pp. 155-167, but only p. 161 is cited by the author of the text.

There are a number of other changes to section 3.11 that should be made for greater internal consistency of the Guidelines and to promote best practice -- some of which were mentioned in the discussion below. See https://sourceforge.net/tracker/?group\_id=106328&atid=644062 .

Original comment by: @kshawkin

[TEITechnicalCouncil]

The URL for changes to 3.11 is supposed to be http://sourceforge.net/tracker/?func=detail&atid=644062&aid=3369109&group\_id=106328 .

Original comment by: @kshawkin

[TEITechnicalCouncil]

Responding to Kevin's last-but-one comment:

This all looks good, but there's one bit that confuses me:

"...if biblScope is a sibling of imprint, it indicates particular volume or page numbers cited by the author.>

I can see that in the case of page numbers, the pages cited are different from the start and end pages of the complete article; but I don't see how there could be a difference between the two <biblScope>s in this fragment:

<monogr> [...] <biblScope type="vol">5</biblScope>

<imprint> <biblScope type="vol">5</biblScope> [...] </imprint>

</monogr>

My question is: How could the two <biblScope>s for "volume" or "issue" be different; and if they were different, what could the difference possibly mean? Surely you can't cite from volume 5 when the article comes from volume 6?

Original comment by: @martindholmes

[TEITechnicalCouncil]

Sorry. Trying to recover from recent events and maybe not fully lucid. How does the point here related to the existence of an analytic sibling to monogr. What would the meaning of biblScope in analytic be (as an alternative to having biblScope as sibling of imprint)?

Original comment by: @laurentromary

[TEITechnicalCouncil]

My original thought was that the <biblScope> which refers to the page numbers of the article within the larger volume should be inside the <analytic> rather than inside the <monogr>; those page numbers are arguably a property of the article rather than the volume. But I could see the other point of view.

So we now have three possible locations for <biblScope>:

1. Inside analytic (where I'd argue the article page numbers should go).

2. As a child of <monogr> (sibling of <imprint>), which is where Kevin says the page number(s) of a specific reference should go.

3. Inside <imprint>.

I would argue that a <biblScope> inside <monogr> or <imprint> could be used to specify the page numbers of an entire journal issue/number within a series, in the case of journals which have sequential page numbers across issues.

Original comment by: @martindholmes

[TEITechnicalCouncil]

I don't recall considering <biblScope> as a child of <analytic> when we prepared the Dublin proposal, but I agree with Martin that the page range of a chapter in a book or article in a journal belongs here rather than in either <monogr> or in <imprint> within <monogr>. So I shouldn't have shut out this option in my comment from 2010-10-22, where I advocated a path most in line with our Dublin proposal which would simply clarify the two encoding options for page numbers and page ranges currently allowed in P5 (which are not well explained in the Guidelines).

Still, if we're trying to put things in sensible places rather than simply clarify the current confusing and not entirely sensible practice, where would you put the particular page(s) quoted or otherwise cited? Perhaps as a direct child of bibl|biblStruct?

If so, I think we could agree on the following:

a) bibl|biblStruct/biblScope[@type='pp'] -- particular page(s) quoted or cited

b) bibl|biblStruct/analytic/biblScope[@type='pp'] -- the page range of a chapter in a book, the page range of an article in a journal, or other page range for the analytic item

c) bibl|biblStruct/monogr/biblScope[@type='pp'] -- the page range of a particular issue of a journal or volume in a monograph series, either of which is described in bibl|biblStruct/series

d) bibl|biblStruct/monogr/imprint/biblScope[@type='pp'] -- no longer recommended

e) bibl|biblStruct/series/biblScope[@type='pp'] -- no longer recommended

Original comment by: @kshawkin

[TEITechnicalCouncil]

This makes sense to me, although I do anticipate some resistance because we're now disrecommending (if that's a word) what was previously the recommended practice (<biblScope> inside <imprint>).

Original comment by: @martindholmes

[TEITechnicalCouncil]

On 11-09-18 11:31 PM, Laurent Romary wrote: > I see the proposal has gone astray and I don't seem to agree with the > direction taken. In any case I don't feel at ease breaking what has > always made sense to me, namely, imprint/biblScope, corresponding to > "the page range of a chapter in a book, the page range of an article > in a journal", which is indeed closely related to the way the article > has been published. > > I actually thought of analytic/biblScope for the quoted pages, and > that's what I was referring to in the ticket. > > Finally, I think biblStruct/biblScope as extremely dangerous, as this > correspond to destructuring a highly structured object such as > biblStruct (and we are not in the context of the bibl-biblScope > debate are we?). > > How do we proceed from this? I would first like to know whether there > are strong diverging opinions among us. Than maybe a dedicated Telco > could be necessary. This is quite an important issue: one > bibliographical standards are set we may quickly get tons of > implemented occurances. We'd better think the thing right.

Original comment by: @kshawkin

[TEITechnicalCouncil]

On 9/19/11 8:44 AM, Martin Holmes wrote: > My opinions are still as they're stated on the ticket: I'm also > reluctant to appear to deprecate imprint/biblScope, just because it's > been so widely used, but I do think analytic/biblScope, monogr/biblScope > and biblStruct/biblScope all have good arguments in their favour, and > particular use-cases. > > However, I no longer really believe that biblStruct is really useful in > itself; I've stopped using it in current projects, in favour of bibl, > now that bibl can be nested. So I'm fairly neutral on whether this > change is made or not.

Original comment by: @kshawkin

[TEITechnicalCouncil]

Let's put aside the question of the usefulness of biblStruct since it's not relevant to this ticket. I am trying to clarify the guidelines on use of biblScope, regardless of whether this is within bibl or biblStruct.

I think there are two separate matters here:

a) where to put the quoted page(s) and where to put the "the page range of a chapter in a book, the page range of an article in a journal".

b) whether, if "the page range of a chapter in a book, the page range of an article in a journal" won't go in imprint/biblScope, whether to deprecate usage of imprint/biblScope in P5

We need to solve (a) before (b). Since we can't reach consensus on (a), I think we're going to need to abandon this ticket. I find that to be a pity since lack of clarify on which page numbers go where hinders interoperability and is a barrier to adoption of the TEI for uses in scholarly publishing.

Original comment by: @kshawkin

[TEITechnicalCouncil]

I would invert the resoning. b) has to remain stable since there are quite some data around using the monogr/imprint locus for biblScope

If we would keep b) stable, we could easily recommend that for a) analytic is the place for quoted pages, keeping the structured nature of biblStruct

Original comment by: @laurentromary

[TEITechnicalCouncil]

That's OK with me. It gives us some progress, at least.

Original comment by: @martindholmes

[TEITechnicalCouncil]

Adding biblScope as a child of analytic seems useful. But if I encountered a biblScope inside a monogr which contains a page range, how would I know if the page range is for whole journal article or just for the cited pages of that article? (The same applies for any analytic work that's not a journal article.)

Original comment by: @kshawkin

[TEITechnicalCouncil]

It should be clear that any biblScope directly within analytic or monogr (don't know what this would mean for series though) should mean the indication of the excerpt. +But+ biblScope in monogr should only be allowed when the isn't an analytic to host this information. So, in the case of a book, one can have such a biblScope in monogr meaning the portion of the bool being quoted, but for a journal paper, the biblScope should exclusively be in analytic (I'm not talking of the one in imprint here to avoid interference)

Original comment by: @laurentromary

[TEITechnicalCouncil]

Laurent is saying that if we made this change to the schema, any <biblScope> within <analytic> which contains a page range should be interpreted as meaning the page range cited, not the page range of the whole analytic item. However, when you are citing from an independent item and not an analytic work -- that is, when there's no <analytic> but only <monogr> in your citation -- you put the page range of the cited pages in a <biblScope> which is a child of <monogr> (or of <imprint> within <monogr>?).

So if I want to indicate the page range of the whole journal article (and not just the part quoted), where do I put this page range? In <monogr>? In <series>? Where do you put the volume and issue number of the cited article?

Original comment by: @kshawkin

[TEITechnicalCouncil]

Further discussion by email among Martin, Laurent, and Kevin led to a unified proposal to deal with the remaining issues. See https://sourceforge.net/tracker/?group\_id=106328&atid=644065 .

Original comment by: @kshawkin

[TEITechnicalCouncil]

• status: open-accepted --> closed-accepted

Original comment by: @kshawkin

[TEITechnicalCouncil]

Argh, I meant to link to https://sourceforge.net/tracker/?func=detail&aid=3497079&group\_id=106328&atid=644065 .

Original comment by: @kshawkin