rfrank has responded, over on the fadedpage forums, to the note i sent yesterday on the posted "sitka" book. as i might have expected, he fell on his sword and took full blame for the errors that were present... in some part, that's correct, because some errors _had_ been reported, and he simply failed to check the forum. but that's not really the proper take-away message here. a proofer can miss things, and still hit the "done" button, just because they honestly felt they'd caught everything. even the best proofer, with an accurate self-perception. 2 of the 6 errors that i reported were cases just like that. if the only person who's gonna serve as "back-up" for that proofer is the post-processor, that puts too much stress on the post-processor. that's exactly what d.p. has done, and that's why they have so few volunteers for that task... rfrank is willing to take on that stress, which is why he has fallen on his sword. but that's the wrong approach to take. rfrank is also improving his tools, so they catch those glitches. that's good, and it's part of the reason i reported these errors. plus he'd already improved his workflow, to catch _italics_... his improved workflow might work, _if_ the o.c.r. recognizes the styling correctly. but on any books with heavy formatting, going back and reinserting the formatting might be a real pain. a better approach, in my view, would be the one that rfrank started with, when he put up his site, which is to encourage the volunteers to do _both_ the proofing and the formatting. it's really _not_ that difficult to do both these tasks together. this is especially true if you give people a _formatted_display_, because then the obtrusive markup is cleared from the screen, and it's replaced by a rendering that resembles the actual scan. i demonstrated this technique with my own proofing site, and showed the additional strength that questionable words can be highlighted in a different color, maximizing the value of a flag. *** rfrank said:

But the most important thing I've concluded is that the majority of reportable errors in Sitka are chargeable to the post-processor (me) and not the roundless system.

see, there's the "experimenter bias" that i was talking about; he'd rather take the blame himself than blame his system... i believe in the roundless system too. i believe in it so much, so strongly, that i believe the evidence can stand up for itself.

We don't have nearly enough participation by a realistic cross-section of typical users here at fadedpage to conclude anything based on real science or real statistics. All I can say is that I am liking the roundless system more and more. It seems to be doing its part well and continues to improve.

there isn't a "cross-section" of "typical" proofers over at fadedpage, it's true. the proofers there are probably much better than average. and, as shown, even these better-than-average proofers can _miss_ errors. we're all human. we make mistakes. we need to be checked. -bowerbird