in "2. Pull up resistors. Make sure RST, GPIO-0, and CD_EN are pulled high." to which pin is that CD_EN? did you mean CH_PD?

on esp-12  RST to high+4.7k doesn't allow to boot, if I disconnect, it does.

If anyone wants to review my code; it's here:

https://codebender.cc/sketch:108178

Even more... So after a lot of experimenting, I'm finding the esp difficult to solidly predict its behavior.  

For example; directly after every 3rd or so server request the unit just decides to reset itself.   I think it has a hard time managing competing requests coming too often.  Gets confused and resets.  It can be detected because the reset leaves the MUX in state 0, and not 1 (AT+CIPMUX?).  I have written fault code to detect this and disconnect and reset on either, but it's troubling.  If you want even a simple yet predictable server; this may not be the device.

Follow up.. I really added a ton of reliability with two simple software/flow fixes.

- Web browsers send two rapid requests per page... The first is for the favicon.  That was happening at a cadence just high enough to cause a lot of weird behavior with overlapping messages coming and going at the same time.  The fix was really easy, just skip the message send and go straight to closing the connection if the request has "/favicon.ico" in it.

- Check values before setting them.  The ESP8266 can hande setting over an existing value, but it often returns odd results that are not predictable and screwup critical timings when it's much much faster to check the value first.

So here's my C# test code for a "simple loop" for being a responsible web server.  I will be translating this to Atmel C++

while (true)
                    {
                        try
                        {
                            if (resetConnection)
                            {
                                if(null != portT && portT.IsOpen)
                                    portT.Close();
                                portT = Init();

                                VerifyBeforeCommand(portT, "CIPMUX", 1);
                                Command(portT,"CIPSERVER",1,80);
                            }

                            portT.ReadTimeout = 60000;
                            portT.ReadTo("+IPD,");
                            portT.ReadTimeout = 500;

                            var stuff = portT.ReadLine();
                            portT.ReadTo("OK");
                            var connectionId = int.Parse(stuff.Split(',').First());

                            if (!stuff.Contains("/favicon.ico"))
                            {
                                var msg = string.Format("

UbiVis-TinyEVSE

Web Controller

Stop Charging", count++);
                                Command(portT, "CIPSEND", connectionId, msg.Length);
                                portT.Write(msg);
                                log.Information("sending message");
                                System.Threading.Thread.Sleep(300);
                                portT.DiscardInBuffer();
                            }

                            CommandOK(portT,"CIPCLOSE", connectionId);
                            log.Information("closed");
                        }
                        catch (ArgumentException) { }
                        catch (TimeoutException) { resetConnection = true; }
                        catch (IOException) { }
                        catch (InvalidOperationException) { }
                    }

I like this post, I too have found many of the same points.  I am attempting to make a stable web server.  Many points of similarity; but some points of differences;

-  3.3v throughout, clocked down some, but everythign off a 3.3v 1A supply.  No levels shifts required.

- I found that reliably flowing the code to ensure MUX and server connection was a must, but I still haven't resolved timing issues.  

Current issues:  It seems to work ok, if I give it enough time to settle and re-establish itself, but when concurrent requests come, it can't handle them and gets lost in the data coming and going and what state it can be in.

ToDos (for me): 

- Setup watchdogs on the device to enhance overall robustness

- Peridically reset the device and rebuild the server in case it's hung.  which seems to happen

Does anyone have good source for a robust server?