I've read several threads here on the subject and am still confused. Some say I should not downsample. "Why throw away pixels?" they say. Some say I should downsample in Lightroom instead of PS because PS does a crappy job at it. And others prescribe special downsampling techniques, like downsampling incrementally several times over.
Lightroom does a better job of down-sampling, with fewer aliasing artifacts than e.g. Photoshop (CS6 and before). There are algorithms that do a slightly better job.
My photo lab's printer's native resolution is 300dpi, but my images' resolution is higher than that (given the print sizes I have planned). [Note: I think my lab is bullshitting me about their Chromira printer's maximum resolution capablilty. They say 300dpi, I say 425dpi. I read somewhere http://www.zbe.com/products/chromira/chromira_5x30/faq
that this printer can go up to 425dpi. But, hey, what do I know. I don't think they want to switch back and forth from 300 to 425 because it will interupt their work schedule. That's the nature of mass printing services, I guess. But their prices are decent.]
Actually their FAQ suggests that their printing technique has the visual effect of a 425 PPI print, because they use a diagonal (diamond) shaped interpolation or maybe just a printing grid. In general, diagonal resolution is indeed a factor of Sqrt(2) higher in a rectangular grid, but the input raster image grid is still 300 PPI (horizontally/vertically). It's probably an internal printer driver recalculation for the printing mechanism that you cannot influence, so you should deliver 300 PPI input files.
If I don't downsample, I'm sure the lab's printer will automatically do it for me; then the sharpening will be all screwed up.
I'd output-sharpen for 300 PPI, do not overdo it as you would for matte inkjet materials but aim for the glossy output setting in Lightroom if you use that, or export unsharpened and do some proper deconvolution sharpening e.g. in Photoshop (using FocusMagic, which is hard to beat, or a similar Plugin) after the resampling to 300 PPI. Just try and make sure that the lab does not use excessive sharpening themselves on top of that (not likely, but better be safe than sorry). You can always ask to print a crop you take out of the full output resampled and sharpened output file, and see how that works out. If you do a test print crop (e.g. 8x10 inch), you can also sharpen half of it, and leave the other half unsharpened, and see how that difference looks in the final product they deliver.
So how exactly should I downsample? Or should I?
With the proper procedures, you will probably be able to get better output quality than they do, because after resampling you can control the output-sharpening much more tuned for the specific image than they will. The only variable is what they will additionally do to your output file.
Resample to 300 PPI, and optionally work on the overall details of the output file with tools like Topaz Detail, or similar, especially if the viewing distance is relatively predictable. Then use deconvolution sharpening to compensate for the slight resampling blur you also created with the resampling (unless you already used output sharpening in Lightroom). Don't over-sharpen to pre-compensate for output losses too much because there will be few, unless the output medium is very textured.