Critic A: "This image has a certain ethereal quality.
Critic B: "It's a fuzzy picture."
The use of soft-focus lenses has always been somewhat controversial as well as rather difficult.  To me, they are a rather specialized tool, something like the fish-eye lenses.  I have seen many fish-eye pictures I didn't care for at all, but a few I found wonderful.  Similarly, there are awful soft-focus images, but some of the all-time great photographs are soft.  The names Stieglitz and Steichen jump into mind. Many think of these lenses as merely a way to "erase wrinkles", but in the right hands they can help in creating beautiful landscapes and provide a mood that strengthens other photographs.  Soft-focus lenses are tools.  Like other tools they can provide results that are good or bad.
No lens is perfect.  All lenses have defects, known as aberrations, that cause the image to be less than perfect.  Softness of focus is the result of one or more of these aberrations being strong enough to be noticeable.  One aberration is the "spherical".  It is comparatively easy to grind a lens surface which is a segment of a sphere, but a surface like that can't quite put an image into correct focus.  Lens designers have to use several surfaces working together to reduce spherical aberration to useful levels.  The most common means of producing a soft-focus lens is to allow some spherical aberration to remain.  A caution.  Some very early soft focus lenses also had significant amounts of chromatic aberration, they couldn't focus different colors onto the same surface.  Unless you intend to use only blue-sensitive film, no longer readily available, these would not be a good idea.  The tip for identifying these lenses is that they will not contain an  achromatic component - two or more glasses together.      Two examples are the Dallmeyer-Bergheim a telephoto soft-focus lens, and the Puligny-Puyo.
Many photographers still alive, and perhaps still working, today, started off with a soft-focus lens, although they probably have never thought of it that way.  During the 1920's and 30's, Box Brownies and other simple cameras were apt to be fitted with a single meniscus lens - one piece of glass with convex curves on both surfaces, thicker in the middle than the edges.  These exhibited most every aberration in the book, but had a small enough aperture that images were usable.
There are two categories of people.  Those who divide into categories, and those who do not.  There are two categories of soft-focus lens.  Those which provide selection of softness without changing the aperture, and those which do not.  let's discuss the latter first.  The smaller the aperture, the less spherical aberration affects the image.  This provides the basis for making the simpler forms of soft-focus lenses.  You need only two pieces of glass to make a basic achromat, a lens which will focus different colors of light together well enough to provide a usable image, but which has too much spherical aberration to produce a sharp image at larger apertures.  Put a diaphragm in front of this lens, stop it down to f16, and you can make a fairly sharp photograph.  This is the construction of two of the four soft-focus lenses that I am aware of being made today.  These are the Rodenstock Imagon and the current version of the Fujinon soft-focus lenses. (Earlier Fujinon SF lenses were triplets.)  These lenses have peculiar diaphragms - removable discs with central holes surrounded by a ring of smaller holes which can be opened or closed.  Buyers of used lenses should be sure that all the discs are present.  Imagon lenses are made in 250 and 300mm focal lengths as well as versions for medium format cameras.  The Fujinons come in 180 and 250mm lengths.  There can be an occasional problem with these lenses.  Strong highlights can be surrounded by a ring of little lights - an image of the diaphragm.
The other two current soft-focus lenses are made by Yamasaki and Cooke.  Under the "Congo" name, soft focus triplets are made by Yamasaki in  150 and 200mm lengths.  To date, Cooke produces only one size, the 9" (229mm) PS945.  The Cooke design is based on a very old lens, the Pinkham-Smith, which, along with the Busch Nicola Perscheid, has become something of a cult item, particularly in Japan, receiving very high prices on the used market.  There were several series of P& S lenses, later sold under the Smith name only, which differed in their characteristics.  I don't know which was used as a design basis by Cooke.
Speaking of used items, most are quite old, but there is less reason to reject a soft-focus lens because of age than there is when searching for a normal lens.  O.K., it won't be as sharp as a new lens.  You don't want it to be.  It won't be coated, and it won't have the contrast of later designs.  Most of your subjects will not want high contrast treatment.  Flare could be a problem, but that's what lens shades and lighting control are for.  Many soft-focus lenses have rather narrow coverage for their focal lengths and this can be helpful where flare is concerned.
Still speaking of lenses which control the amount of softness only by the aperture, there are many different designs which may be available at least occasionally.  One of the more common is the Wollensak Veritar.  They have an achromat pair at the back and a large meniscus lens at the front.  The combination gives a peculiar result in stopping down.  All the gain in depth of field is behind the plane of sharp focus, so instead of focusing on an eye when making a portrait, you need to focus on the tip of the nose.    They are among the few older soft-focus lenses which came in (large) synchronized shutters.  A still earlier Wollensak product was the Verito.  Except for the two smallest sizes, this was an f4 lens, the back group of which could be used by itself.  It may have been a Rapid Rectilinear type, but more likely a Petzval.
The Kodak Portrait Lens is an achromat like those of the Imagon and Fujinon lenses, but has a normal diaphragm instead of removable discs.  The Spencer Port-Land, the Hanovia "Kalosat" and the Cooke Achromatic Portrait are of similar construction.  The same description would also apply  to early "landscape lenses.  The basic difference is that those intended as soft-focus lenses have larger apertures.
Some Dallmeyer lenses fall into this grouping.  One type was just called the "Dallmeyer Soft Focus.  It appears to be similar to the Kodak.  Another was the "Mutac", unusual in that it was a triple convertible.  You could use the lens complete or with either of the cells by itself.
The Petzval Portrait was a special case, not really a soft-focus lens, since it is very sharp in the center.  The very first mathematically computed lens, it goes back to within a year of the public introduction of photography and the aim was to produce a lens fast enough to take people pictures.  Softness in the outer areas was not intentional, but was accepted in order to attain the desired speed.  It also has an inward curving field.  Aside from wide-angles, most of the Darlot lenses found today are modified Petzval types, some, the Ross, for example, offering a flatter field.    Burke & James were still offering new modified Petzval type portrait lenses into the 1970's.
The other category consists of a normal lens, as far as the glass is concerned, with a means of varying the lens elements so as to produce a controllable amount of spherical aberration, thus giving you some measure of control independent of the aperture.  Stopping down will, of course,  still create a sharper image, but you have a degree of control over depth of field in addition to amount of diffusion.
Here we go with categories again.  Some of these lenses are based on anastigmat designs, some are not.  Those which are not are mostly Petzval designs, and those most frequently seen are made by Dallmeyer, who started making portrait lenses with diffusion control in 1866.  You want shallow depth of field?  Try the Dallmeyer Patent Portrait No. 8D, 37" (934mm) focal length, f6.  If speed is your thing, the B series was f3.  Most of these well deserve the term "brass cannon".  You need a very solid camera with a large lensboard to make use of all but the smallest sizes.  The glass ranged up to 6" diameter.  They were also extremely expensive, up to over $400 at a time when you could buy an Eastman View Camera for $19.00.  The smaller sizes carried a rack and pinion focusing movement, and you had to pay extra for an iris diaphragm instead of Waterhouse stops.  Ross also made lenses of this type in f3.5 aperture.
The Wollensak "Vitax" is probably a Petzval type.  It was made in at least three sizes from ten to 16 inches (254 to 406mm).  It is distinctive in having a knob on the side as the diffusion adjustment means.  It may be found in a "Studio" shutter, an iris diaphragm type working in only the "Bulb" mode.  Also apparently a Petzval type is  the Eastman Portrait Lens, not to be confused with the Kodak Portrait of much later production.  "Eagle" Portrait Lenses were sold by George Murphy Inc., and I think this probably was a Murphy house brand.  There were several series, both with and without the diffusion adjustment.
Turning to the anastigmats, many makers provided at least a few lenses with a soft focus adjustment added to lenses selected from their normal production range.  An exception was the Graf "Variable", designed from the start as an adjustable diffusion type.  The name derives from the fact that the focal length and the aperture changed a little as the soft focus was selected.      Edward Weston was among the several prominent users of Graf lenses.  Wollensak made some lenses of this type, probably Tessars.    The ones I have seen were Series II, but there may have been others.  In England, they were made by Taylor, Taylor & Hobson    (Cooke), Dallmeyer, Beck and possibly others.  Cooke lenses are the most common, in Series II, f4.5 and Series VI, f5.6, but not all Series II Cookes have the diffusion device.  Many Cooke lenses have very prominent adjustment handles with two finger openings, sometimes referred to as "spectacles".  The absence of these does not necessarily mean that there is no diffusion adjustment, however.  The only Continental anastigmats with a soft-focus feature that have come to my attention are the Voigtlander "Universal" Heliar and the Zeiss  Portrait Unar, also made under license by Bausch & Lomb.    There probably have been others, though.
Like other tools, the use of soft-focus lenses improves with practice.  There is a problem in that the groundglass gives only a limited idea of what the final print will look like.  It would be a good idea to take a series of otherwise identical photos with differing settings of aperture and (if provided) diffusion control, when first trying out a soft-focus lens.  The resulting prints can then be used as references when making  future photographs.
There are many methods of softening focus without use of a special lens.  One idea, frequently suggested, is to diffuse the focus during enlargement.  There is a commercial item for this purpose, the Pictrol,  too small at 2" inside diameter to go in front of most large format camera lenses.  The problem here is that diffusion during a printing process  produces a different result.  Diffusion onto a negative spreads light out into the shadow areas.  Diffusion during printing spreads darkness into the highlights.  The result has been described as being suitable for portraits of the Addams family.
One early device mechanically jiggled the focus control during exposure, the apparent aim being to increase depth of field.  Some early lenses were advertised as having great depth of field.  This was the same as saying that they were soft.
Some experimenters have used the rather hazardous means of partially unscrewing a cell or cells of a normal lens.  My first photograph with a new wide angle lens produced a picture of a youth choir looking somewhat more angelic than expected.  I traced the phenomenon to a lensboard a little too thick to allow the cells to fully come into position.  As it turned out, the choir director liked the result.    If I were going to try something  that might result in accident to the lens, I think I would try it on one of the surplus Xerox lenses.  These are typically 8 1/4",  f4.5, made by very good manufacturers.  As of this writing, Copy Raptars  are available at www.surplusshed.com at $10.00.  The last I knew, these lenses by T,T& H, B& L and Rank were available  at www.candhsales.com at $16.50.  The Raptars appear to be normal production items, while the others were designed especially for copy machines and may have been color corrected for the near monochromatic light of these machines.
Some people have smeared vaseline onto the front of their lenses.  I think I would prefer to smear up a filter rather than a lens.  Others have used everything from cigar smoke to a piece of ladies stocking material in front of the lens.  In the latter case, sometimes holes are burned into the fabric to modify the results.  Hollywood types have employed fog machines.  There  are commercial diffusing discs.  The Zeiss Softar seems to be the most highly regarded device of this type.  It has a series of concentric thread-like rings formed into the glass and is rather expensive..
There are ways of upsetting carefully calculated spherical aberration reduction besides changing element spacing.  A thick glass plate is one.  Adding various combinations of weak positive and negative elements, Proxar and Distar and the like, might be another.
I have found surprisingly little in the way of compiled information on this subject.  This little article is an attempt to fill this apparent gap.  All comments, suggestions and particularly provisions of additional information are especially welcome.