Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[JohnnyYuma]

Do we have any estimate?

[KittyTac]

Do we have any estimate?

~1-3 months.

[Shonai_Dweller]

July 27th:

The list is down to about twenty nuggets like this, which take a day or two each. Not that I'd extrapolate a release date from that... it never works that way. But we're getting there.

So, ignoring the last bit about guessing being meaningless, and assuming a 1-week 'now into stress testing! Announcement', sometime between September 1st and September 22nd. Assuming Toady doesn't take a day off or report that he has no air conditioning and is feeling 'August heat addlement' and tension from parts of the nearby countryside burning.  Oh wait...

[Toxicshadow]

Every time someone asks the release gets delayed by 3 months

[KittyTac]

Every time someone asks the release gets delayed by 3 months

Dead meme

[Toxicshadow]

Every time someone asks the release gets delayed by 3 months

Dead meme

You're right, I guess I'm a necro then

[King Mir]

Shonai_Dweller quoted last month's hit of an estimate. This month we have an update:

Last month around this time, there were about 20 nuggets left. That number went up and down as work happened and crap popped up -- now there are about a dozen. So progress! Not as much progress as I had hoped, as usual, but not so bad either. I'll keep plugging away at it.

So in 28 days, there were a net 8 nuggets processed, which is 3.5 per day. Extrapolating that pattern, the release will be ready in 6 weeks.

[Shonai_Dweller]

Shonai_Dweller quoted last month's hit of an estimate. This month we have an update:

Last month around this time, there were about 20 nuggets left. That number went up and down as work happened and crap popped up -- now there are about a dozen. So progress! Not as much progress as I had hoped, as usual, but not so bad either. I'll keep plugging away at it.

So in 28 days, there were a net 8 nuggets processed, which is 3.5 per day. Extrapolating that pattern, the release will be ready in 6 weeks.

Or...
One nugget every 2 days means 14 nuggets were cleared, leaving 6. 6 new nuggets appeared meaning there are now 12. That's one new nugget for every 2.33 nuggets solved. So 24 days from now the original 12 will be solved and there'll be 5 brand new nuggets to solve. 10 days after that Toady'll swear at his computer and start the final release anyhow. There'll be time for bug fixing later. So 34 days of nugget bashing, followed by a week of stress testing. That's...6 weeks until ver .01 is released (which will crash for 83% of users). Oh, just what King Mir said.

[King Mir]

So It's unlikely that the one nugget every two days estimate was accurate. The number of nuggets added was therefore probably less than 6. Programmers are notorious for underestimating work required, and ToadyOne is no different.

Also, Shonai_Dweller, your calculation forgets that for the 5 new bugs that turn up, you have to add another 2 or so nuggets discovered while processing those. And then another nugget while processing the 2. That adds a total of 6 days. That would total 40 days, or 40.58 if you use exact fractions the whole calculation, which is much nearer to my 42 day estimate.

It's a good point that there will be some additional testing once all "nuggets" are complete. If your guess of one week is correct, then my estimate would be increased to 7 weeks. Is there reason to believe that final testing takes that long? I thought it was more like a few days.

[Shonai_Dweller]

So It's unlikely that the one nugget every two days estimate was accurate. The number of nuggets added was therefore probably less than 6. Programmers are notorious for underestimating work required, and ToadyOne is no different.

Also, Shonai_Dweller, your calculation forgets that for the 5 new bugs that turn up, you have to add another 2 or so nuggets discovered while processing those. And then another nugget while processing the 2. That adds a total of 6 days. That would total 40 days, or 40.58 if you use exact fractions the whole calculation, which is much nearer to my 42 day estimate.

It's a good point that there will be some additional testing once all "nuggets" are complete. If your guess of one week is correct, then my estimate would be increased to 7 weeks. Is there reason to believe that final testing takes that long? I thought it was more like a few days.

I did add the new nuggets and estimated that at that point Toady wouldn't care any more.

[StagnantSoul]

My guess is the first week of october.

[KittyTac]

Anybody who guesses it right gets a cookie. If I did, I eat it myself.

O

[King Mir]

So It's unlikely that the one nugget every two days estimate was accurate. The number of nuggets added was therefore probably less than 6. Programmers are notorious for underestimating work required, and ToadyOne is no different.

Also, Shonai_Dweller, your calculation forgets that for the 5 new bugs that turn up, you have to add another 2 or so nuggets discovered while processing those. And then another nugget while processing the 2. That adds a total of 6 days. That would total 40 days, or 40.58 if you use exact fractions the whole calculation, which is much nearer to my 42 day estimate.

It's a good point that there will be some additional testing once all "nuggets" are complete. If your guess of one week is correct, then my estimate would be increased to 7 weeks. Is there reason to believe that final testing takes that long? I thought it was more like a few days.

I did add the new nuggets and estimated that at that point Toady wouldn't care any more.

In that case, your choice of when toady stops caring about adding nuggets is rather arbitrary, and might come sooner. Why stop when there's 5 nuggest left, not 10, for example? However, it is a fair point that Toady will probably decrease the rate at which nuggets are added as he gets closer to release.

[Shonai_Dweller]

So It's unlikely that the one nugget every two days estimate was accurate. The number of nuggets added was therefore probably less than 6. Programmers are notorious for underestimating work required, and ToadyOne is no different.

Also, Shonai_Dweller, your calculation forgets that for the 5 new bugs that turn up, you have to add another 2 or so nuggets discovered while processing those. And then another nugget while processing the 2. That adds a total of 6 days. That would total 40 days, or 40.58 if you use exact fractions the whole calculation, which is much nearer to my 42 day estimate.

It's a good point that there will be some additional testing once all "nuggets" are complete. If your guess of one week is correct, then my estimate would be increased to 7 weeks. Is there reason to believe that final testing takes that long? I thought it was more like a few days.

I did add the new nuggets and estimated that at that point Toady wouldn't care any more.

In that case, your choice of when toady stops caring about adding nuggets is rather arbitrary, and might come sooner. Why stop when there's 5 nuggest left, not 10, for example? However, it is a fair point that Toady will probably decrease the rate at which nuggets are added as he gets closer to release.

Yes. As Toady said. It's meaningless to make any calculations (but as fans we love to do so anyhow).

[mikekchar]

TG;IDWR (Too geeky; I don't want to read): I agree with the estimates being given.

There are actually models for "defect discovery rates".  The one I've used the most is called Littlewood's model and although the distribution chosen is likely very naive, I've found that it works reasonably well.  For the academically inclined, http://www.ece.uvic.ca/~itraore/seng426-07/notes/qual07-8.pdf has quite a good introduction.

I like using models that use discovery rates (which Littlewood's model uses) as opposed to defect numbers, although it amounts to the same thing in the end.  The idea is that if you are exercising the code, you will discover defects at a rate that is relative to the density of the defects in the code.  Since arrival rates are usually well modelled with a Poisson distribution, it kind of skips the bit where you have to try to figure out some bizarre distribution (and do calculus with that bizarre distribution).

Like Shonai_Dweller,  I think that the release will be made with known defects.  The key that will trigger release is when the list stops growing (because then the developer thinks, "OK, I can fix whatever defects currently exist in a cleanup release).  Unfortunately, we don't have a published rate of nugget discovery, but luckily delivery times are linearly related to estimates with a normal distribution (BTW, this is probably the most amazingly, amazing thing in software engineering and *nobody* ever writes about it).  Shonai_Dweller's suggestion of 14 nuggets solved with 6 new introduced is probably close enough -- an increase in nugget solving time actually *reduces* discovery rate and makes us *closer* to release, so I think this is a good conservative place to start.

6 new nuggets in 28 days is a discovery rate of 1 nugget per 4.6 days.  We don't need to do any complicated math, really.  As defects are removed, the rate of discovery will go down.  Assuming Toady isn't adding more defects than he's removing (a good bet, I think), then on average we expect it to take more than 4.6 days to discover a new nugget. Let's push it out to 6 days per new nugget (because discovery rate should decrease): that means that there will be 24 days of development and 4 new nuggets.  Crucially, while the last 4 are being developed, odds are that only a single new nugget will be discovered.  I suspect that nugget will not be worked on.

That brings us out to 28 days total.  At that point, 5 days of testing is likely to discover only minor issues that are quick to fix and there will be a release at 33 days out.  I am, however, going to add 5 more days for various things (roughly 1 day a week).  This puts a release at Friday October the 6th.    I think you could probably put error bars out for -5 to +10 days (October 1st is the absolute earliest, while I will be surprised if it goes beyond October 26th).  Interestingly, that is quite different from my "gut" feeling that it will probably come out in November...