Re: SUGGESTION: rollback like GetRight
Dan Harkless <[EMAIL PROTECTED]> writes: > Don't cc me when posting to wget, please. I don't need two copies. > > ZIGLIO Frediano <[EMAIL PROTECTED]> writes: > > In my opinion rollback should be added for two problem: > > - the file change > > - the proxy return unwanted data > > > > The rollback without verify resolve first problem but not second > > The rollback with full verify resolve second but not first > > > > An example: > > G garbage dsata > > D downloaded file correctly > > C download continued > > > > Now you download a file you can obtain something like this (garbage for bad > > proxy): > > G > > In current implementation of wget you finally obtain: > > GCC > > That is a wrong file > > If you rollback before continuing you can obtain > > DDD > > That is correct ... if file didn't change! > > It'd be a lot simpler and would handle almost all continuing-download-of-a- > file-that-has-changed-on-server cases if we made it so if you specify both > -c and -N, wget checks the timestamp on the local file vs. the one on the > server, and if the one on the server is newer, it restarts the download from > scratch. I forgot to add that the timestamp method is superior to the "rollback-verify" method in tons of cases because a file can change without the specific portion you're checking having changed. In any case, I've just added this to the TODO. --- Dan Harkless| To help prevent SPAM contamination, GNU Wget co-maintainer | please do not mention this email http://sunsite.dk/wget/ | address in Usenet posts -- thank you.
Re: SUGGESTION: rollback like GetRight
Don't cc me when posting to wget, please. I don't need two copies. ZIGLIO Frediano <[EMAIL PROTECTED]> writes: > In my opinion rollback should be added for two problem: > - the file change > - the proxy return unwanted data > > The rollback without verify resolve first problem but not second > The rollback with full verify resolve second but not first > > An example: > G garbage dsata > D downloaded file correctly > C download continued > > Now you download a file you can obtain something like this (garbage for bad > proxy): > G > In current implementation of wget you finally obtain: > GCC > That is a wrong file > If you rollback before continuing you can obtain > DDD > That is correct ... if file didn't change! It'd be a lot simpler and would handle almost all continuing-download-of-a- file-that-has-changed-on-server cases if we made it so if you specify both -c and -N, wget checks the timestamp on the local file vs. the one on the server, and if the one on the server is newer, it restarts the download from scratch. --- Dan Harkless| To help prevent SPAM contamination, GNU Wget co-maintainer | please do not mention this email http://sunsite.dk/wget/ | address in Usenet posts -- thank you.
RE: SUGGESTION: rollback like GetRight
> > Jan Prikryl <[EMAIL PROTECTED]> writes: > > Quoting ZIGLIO Frediano ([EMAIL PROTECTED]): > > > I suggest two parameter: > > > - rollback-size > > > - rollback-check-size > > > where 0 <= rollback-check-size <= rollback-size > > > The first for calculate the beginning of range (filesize > - rollback-size) > > > and the second for check (wget should check the range [filesize - > > > rollback-size,filesize - rollback-size + rollback-check-size) ) > > > > My understanding of the rollback problem is that there are > some broken > > proxies that do add some additional text garabge after the conection > > has timed out for example. Then, for `--rollback-size=NUM' after > > timing-out, wget shall cut the last NUM bytes of the file and try to > > resume the download. > > > > Chould you elaborate more on the situation where something like > > `--rollback-check-size' would be needed? What shall be > checked there? > > I think he wants an option where wget will verify that a > certain section > towards the end of the local file matches what's on the > server, so that you > don't have to guess or manually check how far to roll back > by. I don't > think I'd implement it as he's suggesting, though. > True and false. In my opinion rollback should be added for two problem: - the file change - the proxy return unwanted data The rollback without verify resolve first problem but not second The rollback with full verify resolve second but not first An example: G garbage dsata D downloaded file correctly C download continued Now you download a file you can obtain something like this (garbage for bad proxy): G In current implementation of wget you finally obtain: GCC That is a wrong file If you rollback before continuing you can obtain DDD That is correct ... if file didn't change! If change the D part of file is different from server side so you gain another wrong file So I rollback the file DDD(DG) Then start do download DDDCC I can check if CC == (DG). But is always true for garbage data! Now return to file saved: G ... rollback ... DDD(DG) ... start download ... DDDC ... now can check if C == (D). You ignore possible garbage in file. If C == (D) you assume that file is the same. If C != (D) you restart downloading The two parameters should be clearer. rollback-size is the size of (DG), rollback-check-size is the size of (D). freddy77 Entra in www.omnitel.it. Ti aspetta un mondo di servizi on line
Re: SUGGESTION: rollback like GetRight
Jan Prikryl <[EMAIL PROTECTED]> writes: > Quoting ZIGLIO Frediano ([EMAIL PROTECTED]): > > I suggest two parameter: > > - rollback-size > > - rollback-check-size > > where 0 <= rollback-check-size <= rollback-size > > The first for calculate the beginning of range (filesize - rollback-size) > > and the second for check (wget should check the range [filesize - > > rollback-size,filesize - rollback-size + rollback-check-size) ) > > My understanding of the rollback problem is that there are some broken > proxies that do add some additional text garabge after the conection > has timed out for example. Then, for `--rollback-size=NUM' after > timing-out, wget shall cut the last NUM bytes of the file and try to > resume the download. > > Chould you elaborate more on the situation where something like > `--rollback-check-size' would be needed? What shall be checked there? I think he wants an option where wget will verify that a certain section towards the end of the local file matches what's on the server, so that you don't have to guess or manually check how far to roll back by. I don't think I'd implement it as he's suggesting, though. --- Dan Harkless| To help prevent SPAM contamination, GNU Wget co-maintainer | please do not mention this email http://sunsite.dk/wget/ | address in Usenet posts -- thank you.
RE: SUGGESTION: rollback like GetRight
On 10/01/2001 08:50:18 ZIGLIO Frediano wrote: >I suggest two parameter: >- rollback-size >- rollback-check-size >where 0 <= rollback-check-size <= rollback-size >The first for calculate the beginning of range (filesize - rollback-size) >and the second for check (wget should check the range [filesize - >rollback-size,filesize - rollback-size + rollback-check-size) ) > I was thinking of making -c have an optional parameter specifying the rollback. If this was defaulted to 0, it can be given to lseek( , , SEEK_END ) (it would be nice if it could accept a 'k' suffix) The check size then could be specified separately. Csaba -- Csaba Ráduly, Programmer - OS/2 Sophos Anti-Virus email: [EMAIL PROTECTED] http://www.sophos.com US support: +1 888 SOPHOS 9UK Support: +44 1235 559933
RE: SUGGESTION: rollback like GetRight
Rollback is usefull mainly for checking if file is not changed. You check (compare) download data with your file. freddy77 > > Quoting ZIGLIO Frediano ([EMAIL PROTECTED]): > > > I suggest two parameter: > > - rollback-size > > - rollback-check-size > > where 0 <= rollback-check-size <= rollback-size > > The first for calculate the beginning of range (filesize - > rollback-size) > > and the second for check (wget should check the range [filesize - > > rollback-size,filesize - rollback-size + rollback-check-size) ) > > My understanding of the rollback problem is that there are some broken > proxies that do add some additional text garabge after the conection > has timed out for example. Then, for `--rollback-size=NUM' after > timing-out, wget shall cut the last NUM bytes of the file and try to > resume the download. > > Chould you elaborate more on the situation where something like > `--rollback-check-size' would be needed? What shall be checked there? > > -- jan > > +- > - > Jan Prikryl| vr|vis center for virtual reality and > visualisation > <[EMAIL PROTECTED]> | http://www.vrvis.at > +- > - > Entra in www.omnitel.it. Ti aspetta un mondo di servizi on line
Re: SUGGESTION: rollback like GetRight
Quoting ZIGLIO Frediano ([EMAIL PROTECTED]): > I suggest two parameter: > - rollback-size > - rollback-check-size > where 0 <= rollback-check-size <= rollback-size > The first for calculate the beginning of range (filesize - rollback-size) > and the second for check (wget should check the range [filesize - > rollback-size,filesize - rollback-size + rollback-check-size) ) My understanding of the rollback problem is that there are some broken proxies that do add some additional text garabge after the conection has timed out for example. Then, for `--rollback-size=NUM' after timing-out, wget shall cut the last NUM bytes of the file and try to resume the download. Chould you elaborate more on the situation where something like `--rollback-check-size' would be needed? What shall be checked there? -- jan +-- Jan Prikryl| vr|vis center for virtual reality and visualisation <[EMAIL PROTECTED]> | http://www.vrvis.at +--
RE: SUGGESTION: rollback like GetRight
I suggest two parameter: - rollback-size - rollback-check-size where 0 <= rollback-check-size <= rollback-size The first for calculate the beginning of range (filesize - rollback-size) and the second for check (wget should check the range [filesize - rollback-size,filesize - rollback-size + rollback-check-size) ) freddy77 > > Hrvoje Niksic <[EMAIL PROTECTED]> writes: > > Daniel Stenberg <[EMAIL PROTECTED]> writes: > > > > > Could you elaborate on this and describe in what way, > theoretically, > > > the errors would sneak into the destination file? > > > > By a silly proxy inserting a "transfer interrupted" string when the > > transfer between the proxy and the actual server gets interrupted. > > How awful. Okay, I added this to the TODO. I imagine it > won't get done > until someone with one of those broken proxies sends in a > patch to implement > it, though. > > --- > Dan Harkless| To help prevent SPAM contamination, > GNU Wget co-maintainer | please do not mention this email > http://sunsite.dk/wget/ | address in Usenet posts -- thank you. > Entra in www.omnitel.it. Ti aspetta un mondo di servizi on line
Re: SUGGESTION: rollback like GetRight
Hrvoje Niksic <[EMAIL PROTECTED]> writes: > Daniel Stenberg <[EMAIL PROTECTED]> writes: > > > Could you elaborate on this and describe in what way, theoretically, > > the errors would sneak into the destination file? > > By a silly proxy inserting a "transfer interrupted" string when the > transfer between the proxy and the actual server gets interrupted. How awful. Okay, I added this to the TODO. I imagine it won't get done until someone with one of those broken proxies sends in a patch to implement it, though. --- Dan Harkless| To help prevent SPAM contamination, GNU Wget co-maintainer | please do not mention this email http://sunsite.dk/wget/ | address in Usenet posts -- thank you.