For reasons of mechanical efficiency, the Linotype system introduced two-character matrices in 1898.
The two characters were in fact the same character (we now call them glyphs), but in different styles -- usually roman and italic of the same weight, or regular and bold roman.
Therefore, each letter of the font shared its width with the corresponding character of the other font.

And so, many typefaces that were published by Linotype in the early 20th century had identical character widths in regular and bold weights. However, their digital versions might not conform to the original specification.

This equivalence of weight width was useful to typographers in settings such as classified advertising and charts.
Consistency of width across weight is still much used today for tabular figures, with genres such as financial reports in mind.

Hope I got that right, Kent!

BTW, you might find David Consuegra's book useful, in particular his timeline.
http://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Daps&field-keyw...

Thanks Nick. Is it not useful for a typographer to be able to switch weights (other than tabular figures) without having to change things because it doesn't fit anymore? I don't want to base the design of the glyphs on their width. That's why I asked if it's useful to have a font with 'almost' consistent width. Are there any fonts out there that have consistent width through all weights, but don't use the 'two-character matrices' or anything alike (in other words: have a design that is not influenced in shape, only in width, by the concept of consistent width) ?
I hope I'm clear enough.

Take a look at FF Balance: http://www.fontshop.com/fontlist/families/ff_balance/

The Roman weights all have the same width. The Italics have a different width, but again the same in all weights.

Is it not useful for a typographer to be able to switch weights (other than tabular figures) without having to change things because it doesn't fit anymore?

Sure it’s useful, but only under certain circumstances. If a complex document is going to be tweaked a whole lot during the design it’s nice to not have paragraphs reflow. I’m not sure how common a situation like this really is; I don’t see publication designers clamoring for families with matching widths.

If you’re going to do this you should probably plan out a family from scratch and test prototype fonts for the entire family before you undertake designing entire fonts. That way you won’t get to the bold/italic fonts and find that you have to go back and make major changes to the work you already did. Also, it would probably help you to read Stanley Hess’ book The Modification of Letterforms to help you understand how to plan out a type system.

Thanks guys. I just found out Actium has the same width across weights!
@Dunwich: I don't think it's really as hard as you make it sound. The different eights should already have a lot proportions in common, so the differences won't be that big I guess. I might be wrong though. I'm just a roockie after all. To honest I've found most books on typedesign and/or typography quite disappointing, but I'll have a look at the book you suggested. I don't want the italics to get the same width too.

Jos Buivenga has experimented with this:
http://www.josbuivenga.demon.nl/delicious.html

Guardian Egyptian comes with duplexed variants.

The concept of Grades — such as you find in Font Bureau’s Readability Series (Quiosco, for instance) or in some H&FJ workhorses (like Mercury Text, for example) — is built upon consistent widths across all grade variants.

This does not usually extend to the Bold weight, however.

One exception might be Poynter Agate, where I think the consistent widths apply to all weights, as well as all grades, in a style (since an agate type is typically employed in tabular material, like stock listings or sports scores).

If I recall correctly, when Akira Kobayashi created his version of Metro Office (based on Dwiggins’s Metro design), he put all styles on a single set of widths, mimicking the duplexing of metal Linotype designs. I’m not sure this feature has much real value.

Beyond these limited situations, I’m not certain about any real practical value to consistent widths across a broad weight spectrum. Seems like a lot of sacrifice might be necessary. And I can’t think of any practical value to having them “close but not quite,” unless it just happens that way.

But maybe it would be an interesting gimmick. The proof will be in the design.

Here is an example of the typical selection of two-letter matrices for a Linotype face:

As can be seen, the usual case is to put the regular form of the face on the matrix with the italic. Small caps are on two-letter matrices as well, but with the numerals and ligatures, so sometimes a two-letter matrix can have two different characters on it, instead of two different styles of the same character.

However, it is true that bold sometimes also shared the matrix with the normal weight:

because there were purposes for which bold, rather than italic, would be mixed with the text.

Italics having the same width as plain is almost a defining characteristic of Linotype faces; as setting requiring roman, italic and bold is also common, I think some faces also had a wider bold on one-letter matrices that were put into an attachment to the Linotype machine to allow it to have a wider selection of characters.

Else

Struggling to find a copy of The Modification of Letterforms.

There's a few more mentioned in this thread.

I had missed this. FWIW, I call this "unwidth". I think it's definitely an advantage... as long as it doesn't cause more important things to go off. It's good because if you want to change the style of even just a snippet of your text after you've made everything look nice it can make all the difference. There are some other fringe advantages too, for example in some kinds of animated text.

The problem is that beyond a certain weight range the character of the design -arguably more important than the advantages above- goes off. Looking at Else* I have to say that the -resultant- width difference between the Light and the Bold is too much. But strangely enough in Axia** it doesn't seem so bad... It might have to do with how strong the character of the design is: Else is pretty generic, so it can't handle the width difference, while Axia is pretty distinctive so the width difference doesn't overpower its character. Maybe.

* http://www.myfonts.com/fonts/adobe/else-npl/
** http://typographica.org/typeface-reviews/axia/

Another thing that might suffer too much between very different weights is apparent size; for example you rarely want a passage in Bold looking smaller than a Light it's embedded in or next to. Look how much bigger Else's Light looks than its Bold.

BTW Jasper, I see no advantage to "almost uniwidth".

See also:
http://typophile.com/node/91519
http://www.typophile.com/node/61252

hhp

More here:
http://typophile.com/node/100708

Oh yeah - thanks. :-)

BTW I've previously mentioned what I consider to be the "best compromise solution" to the problem I stated above, but this is a good place to repeat it: make the cluster of weights that are pretty close to the Regular uniwidth, and make the outlying weights a fixed tracking amount away from that.* In this way the design of the outlying weights isn't notably compromised, but if the user has a strong need to maintain linebreaks (or wants to do some overlay/fading animation) he can decide to sacrifice the spacing a bit to achieve that. I've been calling this "fixed offset" spacing.

* I've done that in Nour&Patria, where the Regular and Demi are uniwidth, while the Light/Bold are +10/-10 tracking units away from them. BTW Nina and I almost managed to pull it off for Ernestine(&Vem) where the difference between each step was going to be 13 units, but it became too much of a time hog.

hhp